{
    "segments": [
        {
            "id": 1,
            "seek": 2620,
            "start": 5.79,
            "end": 11.75,
            "text": " Hi, and today with Young-Hoo Lee, I'll be presenting Detecting Handcrafted Social",
            "tokens": [
                50404,
                2421,
                11,
                293,
                965,
                365,
                8160,
                12,
                39,
                1986,
                6957,
                11,
                286,
                603,
                312,
                15578,
                4237,
                557,
                278,
                8854,
                5611,
                292,
                9909,
                50700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2963655037076577,
            "compression_ratio": 1.5702127659574467,
            "no_speech_prob": 0.2744140625,
            "words": [
                {
                    "start": 5.79,
                    "end": 6.09,
                    "word": " Hi",
                    "probability": 0.08209228515625
                },
                {
                    "start": 6.09,
                    "end": 6.35,
                    "word": ",",
                    "probability": 0.59326171875
                },
                {
                    "start": 6.35,
                    "end": 6.55,
                    "word": " and",
                    "probability": 0.3310546875
                },
                {
                    "start": 6.55,
                    "end": 7.51,
                    "word": " today",
                    "probability": 0.77001953125
                },
                {
                    "start": 7.51,
                    "end": 7.71,
                    "word": " with",
                    "probability": 0.72119140625
                },
                {
                    "start": 7.71,
                    "end": 8.53,
                    "word": " Young",
                    "probability": 0.197509765625
                },
                {
                    "start": 8.53,
                    "end": 8.71,
                    "word": "-Hoo",
                    "probability": 0.5921630859375
                },
                {
                    "start": 8.71,
                    "end": 8.95,
                    "word": " Lee",
                    "probability": 0.82861328125
                },
                {
                    "start": 8.95,
                    "end": 9.19,
                    "word": ",",
                    "probability": 0.62744140625
                },
                {
                    "start": 9.19,
                    "end": 9.45,
                    "word": " I'll",
                    "probability": 0.94482421875
                },
                {
                    "start": 9.45,
                    "end": 9.65,
                    "word": " be",
                    "probability": 0.9052734375
                },
                {
                    "start": 9.65,
                    "end": 10.09,
                    "word": " presenting",
                    "probability": 0.6806640625
                },
                {
                    "start": 10.09,
                    "end": 10.65,
                    "word": " Detecting",
                    "probability": 0.7194417317708334
                },
                {
                    "start": 10.65,
                    "end": 11.45,
                    "word": " Handcrafted",
                    "probability": 0.8671875
                },
                {
                    "start": 11.45,
                    "end": 11.75,
                    "word": " Social",
                    "probability": 0.77978515625
                }
            ]
        },
        {
            "id": 2,
            "seek": 2620,
            "start": 11.75,
            "end": 15.51,
            "text": " Engineering Emails with a Bleeding Edge Neural Language Model.",
            "tokens": [
                50700,
                16215,
                3968,
                6227,
                365,
                257,
                30721,
                9794,
                19328,
                1734,
                1807,
                24445,
                17105,
                13,
                50856
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2963655037076577,
            "compression_ratio": 1.5702127659574467,
            "no_speech_prob": 0.2744140625,
            "words": [
                {
                    "start": 11.75,
                    "end": 12.29,
                    "word": " Engineering",
                    "probability": 0.8466796875
                },
                {
                    "start": 12.29,
                    "end": 12.81,
                    "word": " Emails",
                    "probability": 0.859619140625
                },
                {
                    "start": 12.81,
                    "end": 13.23,
                    "word": " with",
                    "probability": 0.740234375
                },
                {
                    "start": 13.37,
                    "end": 13.61,
                    "word": " a Bleeding",
                    "probability": 0.927978515625
                },
                {
                    "start": 13.61,
                    "end": 13.85,
                    "word": " Edge",
                    "probability": 0.498046875
                },
                {
                    "start": 13.85,
                    "end": 14.21,
                    "word": " Neural",
                    "probability": 0.946044921875
                },
                {
                    "start": 14.21,
                    "end": 14.55,
                    "word": " Language",
                    "probability": 0.7939453125
                },
                {
                    "start": 14.55,
                    "end": 14.91,
                    "word": " Model",
                    "probability": 0.74951171875
                },
                {
                    "start": 14.91,
                    "end": 15.51,
                    "word": ".",
                    "probability": 0.9306640625
                }
            ]
        },
        {
            "id": 3,
            "seek": 2620,
            "start": 19.94,
            "end": 22.7,
            "text": " A bit about us before getting into the talk.",
            "tokens": [
                50936,
                316,
                857,
                466,
                505,
                949,
                1242,
                666,
                264,
                751,
                13,
                51028
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2963655037076577,
            "compression_ratio": 1.5702127659574467,
            "no_speech_prob": 0.2744140625,
            "words": [
                {
                    "start": 19.94,
                    "end": 20.54,
                    "word": " A",
                    "probability": 0.87939453125
                },
                {
                    "start": 20.54,
                    "end": 20.64,
                    "word": " bit",
                    "probability": 0.9013671875
                },
                {
                    "start": 20.64,
                    "end": 20.84,
                    "word": " about",
                    "probability": 0.826171875
                },
                {
                    "start": 20.84,
                    "end": 21.12,
                    "word": " us",
                    "probability": 0.90625
                },
                {
                    "start": 21.12,
                    "end": 21.36,
                    "word": " before",
                    "probability": 0.71142578125
                },
                {
                    "start": 21.36,
                    "end": 21.62,
                    "word": " getting",
                    "probability": 0.8837890625
                },
                {
                    "start": 21.62,
                    "end": 21.8,
                    "word": " into",
                    "probability": 0.72265625
                },
                {
                    "start": 21.8,
                    "end": 21.94,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 21.94,
                    "end": 22.24,
                    "word": " talk",
                    "probability": 0.82763671875
                },
                {
                    "start": 22.24,
                    "end": 22.7,
                    "word": ".",
                    "probability": 0.640625
                }
            ]
        },
        {
            "id": 4,
            "seek": 2620,
            "start": 22.7,
            "end": 28.68,
            "text": " Sophos is a security products and services company. We have a whole range of security",
            "tokens": [
                51140,
                18921,
                329,
                307,
                257,
                3825,
                3383,
                293,
                3328,
                2237,
                13,
                492,
                362,
                257,
                1379,
                3613,
                295,
                3825,
                51356
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2963655037076577,
            "compression_ratio": 1.5702127659574467,
            "no_speech_prob": 0.2744140625,
            "words": [
                {
                    "start": 22.7,
                    "end": 23.04,
                    "word": " Sophos",
                    "probability": 0.613525390625
                },
                {
                    "start": 23.04,
                    "end": 24.94,
                    "word": " is",
                    "probability": 0.8662109375
                },
                {
                    "start": 25.1,
                    "end": 25.46,
                    "word": " a security",
                    "probability": 0.89453125
                },
                {
                    "start": 25.46,
                    "end": 26.0,
                    "word": " products",
                    "probability": 0.69873046875
                },
                {
                    "start": 26.0,
                    "end": 26.2,
                    "word": " and",
                    "probability": 0.89208984375
                },
                {
                    "start": 26.2,
                    "end": 26.5,
                    "word": " services",
                    "probability": 0.88916015625
                },
                {
                    "start": 26.5,
                    "end": 26.9,
                    "word": " company",
                    "probability": 0.81884765625
                },
                {
                    "start": 26.9,
                    "end": 27.36,
                    "word": ".",
                    "probability": 0.90234375
                },
                {
                    "start": 27.36,
                    "end": 27.5,
                    "word": " We",
                    "probability": 0.9033203125
                },
                {
                    "start": 27.5,
                    "end": 27.78,
                    "word": " have",
                    "probability": 0.88134765625
                },
                {
                    "start": 27.88,
                    "end": 28.0,
                    "word": " a whole",
                    "probability": 0.80859375
                },
                {
                    "start": 28.0,
                    "end": 28.24,
                    "word": " range",
                    "probability": 0.794921875
                },
                {
                    "start": 28.24,
                    "end": 28.36,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 28.36,
                    "end": 28.68,
                    "word": " security",
                    "probability": 0.92431640625
                }
            ]
        },
        {
            "id": 5,
            "seek": 2620,
            "start": 28.68,
            "end": 35.02,
            "text": " products ranging from firewalls, to mobile security, to email security, to endpoint security.",
            "tokens": [
                51356,
                3383,
                25532,
                490,
                36109,
                82,
                11,
                281,
                6013,
                3825,
                11,
                281,
                3796,
                3825,
                11,
                281,
                35795,
                3825,
                13,
                51644
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2963655037076577,
            "compression_ratio": 1.5702127659574467,
            "no_speech_prob": 0.2744140625,
            "words": [
                {
                    "start": 28.68,
                    "end": 29.12,
                    "word": " products",
                    "probability": 0.7744140625
                },
                {
                    "start": 29.12,
                    "end": 29.44,
                    "word": " ranging",
                    "probability": 0.6826171875
                },
                {
                    "start": 29.44,
                    "end": 29.9,
                    "word": " from",
                    "probability": 0.82080078125
                },
                {
                    "start": 29.9,
                    "end": 30.78,
                    "word": " firewalls",
                    "probability": 0.792236328125
                },
                {
                    "start": 30.78,
                    "end": 30.82,
                    "word": ",",
                    "probability": 0.2266845703125
                },
                {
                    "start": 30.82,
                    "end": 30.92,
                    "word": " to",
                    "probability": 0.8955078125
                },
                {
                    "start": 30.92,
                    "end": 31.1,
                    "word": " mobile",
                    "probability": 0.771484375
                },
                {
                    "start": 31.1,
                    "end": 31.5,
                    "word": " security",
                    "probability": 0.9189453125
                },
                {
                    "start": 31.5,
                    "end": 31.64,
                    "word": ",",
                    "probability": 0.95458984375
                },
                {
                    "start": 31.64,
                    "end": 31.72,
                    "word": " to",
                    "probability": 0.87841796875
                },
                {
                    "start": 31.72,
                    "end": 31.9,
                    "word": " email",
                    "probability": 0.87548828125
                },
                {
                    "start": 31.9,
                    "end": 32.4,
                    "word": " security",
                    "probability": 0.9189453125
                },
                {
                    "start": 32.4,
                    "end": 33.1,
                    "word": ",",
                    "probability": 0.9287109375
                },
                {
                    "start": 33.1,
                    "end": 33.4,
                    "word": " to",
                    "probability": 0.89501953125
                },
                {
                    "start": 33.4,
                    "end": 33.86,
                    "word": " endpoint",
                    "probability": 0.6123046875
                },
                {
                    "start": 33.86,
                    "end": 34.42,
                    "word": " security",
                    "probability": 0.921875
                },
                {
                    "start": 34.42,
                    "end": 35.02,
                    "word": ".",
                    "probability": 0.7275390625
                }
            ]
        },
        {
            "id": 6,
            "seek": 5464,
            "start": 35.02,
            "end": 43.04,
            "text": " And the team that I manage, Sophos AI, is the AI shop inside of Sophos. We research and develop",
            "tokens": [
                50408,
                400,
                264,
                1469,
                300,
                286,
                3067,
                11,
                18921,
                329,
                7318,
                11,
                307,
                264,
                7318,
                3945,
                1854,
                295,
                18921,
                329,
                13,
                492,
                2132,
                293,
                1499,
                50768
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2869318200783296,
            "compression_ratio": 1.6567164179104477,
            "no_speech_prob": 0.061767578125,
            "words": [
                {
                    "start": 35.02,
                    "end": 35.74,
                    "word": " And",
                    "probability": 0.1285400390625
                },
                {
                    "start": 35.74,
                    "end": 36.64,
                    "word": " the",
                    "probability": 0.77197265625
                },
                {
                    "start": 36.64,
                    "end": 36.84,
                    "word": " team",
                    "probability": 0.91748046875
                },
                {
                    "start": 36.84,
                    "end": 36.94,
                    "word": " that",
                    "probability": 0.86962890625
                },
                {
                    "start": 36.94,
                    "end": 37.06,
                    "word": " I",
                    "probability": 0.9873046875
                },
                {
                    "start": 37.06,
                    "end": 37.42,
                    "word": " manage",
                    "probability": 0.7177734375
                },
                {
                    "start": 37.42,
                    "end": 37.64,
                    "word": ",",
                    "probability": 0.56982421875
                },
                {
                    "start": 37.64,
                    "end": 37.86,
                    "word": " Sophos",
                    "probability": 0.938720703125
                },
                {
                    "start": 37.86,
                    "end": 38.1,
                    "word": " AI",
                    "probability": 0.7578125
                },
                {
                    "start": 38.1,
                    "end": 38.38,
                    "word": ",",
                    "probability": 0.779296875
                },
                {
                    "start": 38.38,
                    "end": 38.74,
                    "word": " is",
                    "probability": 0.90087890625
                },
                {
                    "start": 38.74,
                    "end": 39.04,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 39.04,
                    "end": 39.32,
                    "word": " AI",
                    "probability": 0.8525390625
                },
                {
                    "start": 39.32,
                    "end": 39.78,
                    "word": " shop",
                    "probability": 0.884765625
                },
                {
                    "start": 39.78,
                    "end": 40.2,
                    "word": " inside",
                    "probability": 0.8759765625
                },
                {
                    "start": 40.2,
                    "end": 40.46,
                    "word": " of",
                    "probability": 0.880859375
                },
                {
                    "start": 40.46,
                    "end": 41.1,
                    "word": " Sophos",
                    "probability": 0.8251953125
                },
                {
                    "start": 41.1,
                    "end": 41.36,
                    "word": ".",
                    "probability": 0.9375
                },
                {
                    "start": 41.36,
                    "end": 41.72,
                    "word": " We",
                    "probability": 0.89453125
                },
                {
                    "start": 41.72,
                    "end": 42.48,
                    "word": " research",
                    "probability": 0.72216796875
                },
                {
                    "start": 42.48,
                    "end": 42.68,
                    "word": " and",
                    "probability": 0.8955078125
                },
                {
                    "start": 42.68,
                    "end": 43.04,
                    "word": " develop",
                    "probability": 0.6611328125
                }
            ]
        },
        {
            "id": 7,
            "seek": 5464,
            "start": 43.04,
            "end": 47.0,
            "text": " the company's machine learning technology. And then we're also responsible...",
            "tokens": [
                50768,
                264,
                2237,
                311,
                3479,
                2539,
                2899,
                13,
                400,
                550,
                321,
                434,
                611,
                6250,
                485,
                50952
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2869318200783296,
            "compression_ratio": 1.6567164179104477,
            "no_speech_prob": 0.061767578125,
            "words": [
                {
                    "start": 43.04,
                    "end": 43.44,
                    "word": " the",
                    "probability": 0.79736328125
                },
                {
                    "start": 43.44,
                    "end": 44.04,
                    "word": " company's",
                    "probability": 0.87939453125
                },
                {
                    "start": 44.04,
                    "end": 44.36,
                    "word": " machine",
                    "probability": 0.708984375
                },
                {
                    "start": 44.36,
                    "end": 44.58,
                    "word": " learning",
                    "probability": 0.81396484375
                },
                {
                    "start": 44.58,
                    "end": 45.16,
                    "word": " technology",
                    "probability": 0.89306640625
                },
                {
                    "start": 45.16,
                    "end": 45.72,
                    "word": ".",
                    "probability": 0.483642578125
                },
                {
                    "start": 45.76,
                    "end": 45.76,
                    "word": " And",
                    "probability": 0.45556640625
                },
                {
                    "start": 45.76,
                    "end": 45.86,
                    "word": " then",
                    "probability": 0.32275390625
                },
                {
                    "start": 45.86,
                    "end": 45.96,
                    "word": " we're",
                    "probability": 0.8916015625
                },
                {
                    "start": 45.96,
                    "end": 46.1,
                    "word": " also",
                    "probability": 0.81201171875
                },
                {
                    "start": 46.1,
                    "end": 47.0,
                    "word": " responsible...",
                    "probability": 0.39898681640625
                }
            ]
        },
        {
            "id": 8,
            "seek": 5464,
            "start": 47.0,
            "end": 50.34,
            "text": " Okay, I'm gonna mute myself. I'm gonna like definitely mute myself in here so I can listen",
            "tokens": [
                50952,
                1033,
                11,
                286,
                478,
                799,
                24523,
                2059,
                13,
                286,
                478,
                799,
                411,
                2138,
                24523,
                2059,
                294,
                510,
                370,
                286,
                393,
                2140,
                51124
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2869318200783296,
            "compression_ratio": 1.6567164179104477,
            "no_speech_prob": 0.061767578125,
            "words": [
                {
                    "start": 47.0,
                    "end": 47.1,
                    "word": " Okay",
                    "probability": 0.241455078125
                },
                {
                    "start": 47.1,
                    "end": 47.2,
                    "word": ",",
                    "probability": 0.6923828125
                },
                {
                    "start": 47.2,
                    "end": 47.32,
                    "word": " I'm",
                    "probability": 0.99169921875
                },
                {
                    "start": 47.32,
                    "end": 47.42,
                    "word": " gonna",
                    "probability": 0.39990234375
                },
                {
                    "start": 47.42,
                    "end": 47.62,
                    "word": " mute",
                    "probability": 0.740234375
                },
                {
                    "start": 47.62,
                    "end": 47.94,
                    "word": " myself",
                    "probability": 0.71728515625
                },
                {
                    "start": 47.94,
                    "end": 48.04,
                    "word": ".",
                    "probability": 0.5283203125
                },
                {
                    "start": 48.04,
                    "end": 48.14,
                    "word": " I'm",
                    "probability": 0.98291015625
                },
                {
                    "start": 48.14,
                    "end": 48.24,
                    "word": " gonna",
                    "probability": 0.69921875
                },
                {
                    "start": 48.24,
                    "end": 48.6,
                    "word": " like",
                    "probability": 0.364990234375
                },
                {
                    "start": 48.6,
                    "end": 48.88,
                    "word": " definitely",
                    "probability": 0.20849609375
                },
                {
                    "start": 48.88,
                    "end": 49.18,
                    "word": " mute",
                    "probability": 0.8232421875
                },
                {
                    "start": 49.18,
                    "end": 49.52,
                    "word": " myself",
                    "probability": 0.69677734375
                },
                {
                    "start": 49.52,
                    "end": 49.7,
                    "word": " in",
                    "probability": 0.87548828125
                },
                {
                    "start": 49.7,
                    "end": 49.82,
                    "word": " here",
                    "probability": 0.8115234375
                },
                {
                    "start": 49.82,
                    "end": 49.98,
                    "word": " so",
                    "probability": 0.7236328125
                },
                {
                    "start": 49.98,
                    "end": 50.06,
                    "word": " I",
                    "probability": 0.98876953125
                },
                {
                    "start": 50.06,
                    "end": 50.16,
                    "word": " can",
                    "probability": 0.88818359375
                },
                {
                    "start": 50.16,
                    "end": 50.34,
                    "word": " listen",
                    "probability": 0.79345703125
                }
            ]
        },
        {
            "id": 9,
            "seek": 5464,
            "start": 50.34,
            "end": 56.18,
            "text": " to the talk. I'm happy if you need something. Young-Hoo Lee is one of the star researchers",
            "tokens": [
                51124,
                281,
                264,
                751,
                13,
                286,
                478,
                2055,
                498,
                291,
                643,
                746,
                13,
                8160,
                12,
                39,
                1986,
                6957,
                307,
                472,
                295,
                264,
                3543,
                10309,
                51436
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2869318200783296,
            "compression_ratio": 1.6567164179104477,
            "no_speech_prob": 0.061767578125,
            "words": [
                {
                    "start": 50.34,
                    "end": 50.44,
                    "word": " to",
                    "probability": 0.884765625
                },
                {
                    "start": 50.44,
                    "end": 50.54,
                    "word": " the",
                    "probability": 0.79736328125
                },
                {
                    "start": 50.54,
                    "end": 50.74,
                    "word": " talk",
                    "probability": 0.8095703125
                },
                {
                    "start": 50.74,
                    "end": 51.3,
                    "word": ".",
                    "probability": 0.9267578125
                },
                {
                    "start": 52.0,
                    "end": 52.46,
                    "word": " I'm",
                    "probability": 0.4755859375
                },
                {
                    "start": 52.46,
                    "end": 53.3,
                    "word": " happy",
                    "probability": 0.693359375
                },
                {
                    "start": 53.3,
                    "end": 53.5,
                    "word": " if",
                    "probability": 0.88330078125
                },
                {
                    "start": 53.5,
                    "end": 53.58,
                    "word": " you",
                    "probability": 0.89453125
                },
                {
                    "start": 53.58,
                    "end": 53.7,
                    "word": " need",
                    "probability": 0.8115234375
                },
                {
                    "start": 53.7,
                    "end": 53.9,
                    "word": " something",
                    "probability": 0.6962890625
                },
                {
                    "start": 53.9,
                    "end": 54.0,
                    "word": ".",
                    "probability": 0.875
                },
                {
                    "start": 54.0,
                    "end": 54.0,
                    "word": " Young",
                    "probability": 0.0002315044403076172
                },
                {
                    "start": 54.0,
                    "end": 54.0,
                    "word": "-Hoo",
                    "probability": 0.015244801839192709
                },
                {
                    "start": 54.0,
                    "end": 54.02,
                    "word": " Lee",
                    "probability": 0.049285888671875
                },
                {
                    "start": 54.02,
                    "end": 54.26,
                    "word": " is",
                    "probability": 0.888671875
                },
                {
                    "start": 54.26,
                    "end": 54.56,
                    "word": " one",
                    "probability": 0.771484375
                },
                {
                    "start": 54.56,
                    "end": 54.7,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 54.7,
                    "end": 55.2,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 55.2,
                    "end": 55.78,
                    "word": " star",
                    "probability": 0.86083984375
                },
                {
                    "start": 55.78,
                    "end": 56.18,
                    "word": " researchers",
                    "probability": 0.76318359375
                }
            ]
        },
        {
            "id": 10,
            "seek": 5464,
            "start": 56.18,
            "end": 63.46,
            "text": " on my team. He's almost completely solely responsible for the machine learning subsystem",
            "tokens": [
                51436,
                322,
                452,
                1469,
                13,
                634,
                311,
                1920,
                2584,
                23309,
                6250,
                337,
                264,
                3479,
                2539,
                2090,
                9321,
                51800
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2869318200783296,
            "compression_ratio": 1.6567164179104477,
            "no_speech_prob": 0.061767578125,
            "words": [
                {
                    "start": 56.18,
                    "end": 56.48,
                    "word": " on",
                    "probability": 0.9072265625
                },
                {
                    "start": 56.48,
                    "end": 56.8,
                    "word": " my",
                    "probability": 0.89501953125
                },
                {
                    "start": 56.8,
                    "end": 57.08,
                    "word": " team",
                    "probability": 0.92236328125
                },
                {
                    "start": 57.08,
                    "end": 57.4,
                    "word": ".",
                    "probability": 0.96044921875
                },
                {
                    "start": 57.4,
                    "end": 57.86,
                    "word": " He's",
                    "probability": 0.938720703125
                },
                {
                    "start": 57.86,
                    "end": 58.98,
                    "word": " almost",
                    "probability": 0.65283203125
                },
                {
                    "start": 58.98,
                    "end": 60.08,
                    "word": " completely",
                    "probability": 0.53564453125
                },
                {
                    "start": 60.08,
                    "end": 61.12,
                    "word": " solely",
                    "probability": 0.52734375
                },
                {
                    "start": 61.12,
                    "end": 61.62,
                    "word": " responsible",
                    "probability": 0.79931640625
                },
                {
                    "start": 61.62,
                    "end": 61.96,
                    "word": " for",
                    "probability": 0.8955078125
                },
                {
                    "start": 61.96,
                    "end": 62.34,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 62.34,
                    "end": 62.66,
                    "word": " machine",
                    "probability": 0.6865234375
                },
                {
                    "start": 62.66,
                    "end": 62.88,
                    "word": " learning",
                    "probability": 0.82373046875
                },
                {
                    "start": 62.88,
                    "end": 63.46,
                    "word": " subsystem",
                    "probability": 0.887451171875
                }
            ]
        },
        {
            "id": 11,
            "seek": 7918,
            "start": 63.46,
            "end": 72.18,
            "text": " that protects our Android customers, at least responsible for the R&D part of that work.",
            "tokens": [
                50376,
                300,
                22583,
                527,
                8853,
                4581,
                11,
                412,
                1935,
                6250,
                337,
                264,
                497,
                5,
                35,
                644,
                295,
                300,
                589,
                13,
                50796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22120786784739976,
            "compression_ratio": 1.5851528384279476,
            "no_speech_prob": 0.0003299713134765625,
            "words": [
                {
                    "start": 63.46,
                    "end": 63.98,
                    "word": " that",
                    "probability": 0.240234375
                },
                {
                    "start": 63.98,
                    "end": 64.42,
                    "word": " protects",
                    "probability": 0.06787109375
                },
                {
                    "start": 64.42,
                    "end": 64.74,
                    "word": " our",
                    "probability": 0.7998046875
                },
                {
                    "start": 64.74,
                    "end": 65.12,
                    "word": " Android",
                    "probability": 0.853515625
                },
                {
                    "start": 65.12,
                    "end": 65.8,
                    "word": " customers",
                    "probability": 0.78173828125
                },
                {
                    "start": 65.8,
                    "end": 67.32,
                    "word": ",",
                    "probability": 0.52197265625
                },
                {
                    "start": 67.32,
                    "end": 67.7,
                    "word": " at",
                    "probability": 0.67724609375
                },
                {
                    "start": 67.7,
                    "end": 67.98,
                    "word": " least",
                    "probability": 0.849609375
                },
                {
                    "start": 67.98,
                    "end": 69.54,
                    "word": " responsible",
                    "probability": 0.30419921875
                },
                {
                    "start": 69.54,
                    "end": 69.96,
                    "word": " for",
                    "probability": 0.88720703125
                },
                {
                    "start": 69.96,
                    "end": 70.6,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 70.6,
                    "end": 70.86,
                    "word": " R",
                    "probability": 0.94677734375
                },
                {
                    "start": 70.86,
                    "end": 71.08,
                    "word": "&D",
                    "probability": 0.89794921875
                },
                {
                    "start": 71.08,
                    "end": 71.38,
                    "word": " part",
                    "probability": 0.82568359375
                },
                {
                    "start": 71.38,
                    "end": 71.46,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 71.46,
                    "end": 71.62,
                    "word": " that",
                    "probability": 0.8818359375
                },
                {
                    "start": 71.62,
                    "end": 71.92,
                    "word": " work",
                    "probability": 0.806640625
                },
                {
                    "start": 71.92,
                    "end": 72.18,
                    "word": ".",
                    "probability": 0.8876953125
                }
            ]
        },
        {
            "id": 12,
            "seek": 7918,
            "start": 72.18,
            "end": 76.0,
            "text": " He's also the principal author of the work presented today. I'm presenting in a supporting",
            "tokens": [
                50796,
                634,
                311,
                611,
                264,
                9716,
                3793,
                295,
                264,
                589,
                8212,
                965,
                13,
                286,
                478,
                15578,
                294,
                257,
                7231,
                50996
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22120786784739976,
            "compression_ratio": 1.5851528384279476,
            "no_speech_prob": 0.0003299713134765625,
            "words": [
                {
                    "start": 72.18,
                    "end": 72.5,
                    "word": " He's",
                    "probability": 0.904052734375
                },
                {
                    "start": 72.5,
                    "end": 72.64,
                    "word": " also",
                    "probability": 0.8291015625
                },
                {
                    "start": 72.64,
                    "end": 72.82,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 72.82,
                    "end": 73.02,
                    "word": " principal",
                    "probability": 0.55078125
                },
                {
                    "start": 73.02,
                    "end": 73.28,
                    "word": " author",
                    "probability": 0.8193359375
                },
                {
                    "start": 73.28,
                    "end": 73.44,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 73.44,
                    "end": 73.54,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 73.54,
                    "end": 73.7,
                    "word": " work",
                    "probability": 0.82568359375
                },
                {
                    "start": 73.7,
                    "end": 74.0,
                    "word": " presented",
                    "probability": 0.6943359375
                },
                {
                    "start": 74.0,
                    "end": 74.38,
                    "word": " today",
                    "probability": 0.79345703125
                },
                {
                    "start": 74.38,
                    "end": 74.78,
                    "word": ".",
                    "probability": 0.95751953125
                },
                {
                    "start": 74.78,
                    "end": 75.18,
                    "word": " I'm",
                    "probability": 0.9892578125
                },
                {
                    "start": 75.18,
                    "end": 75.44,
                    "word": " presenting",
                    "probability": 0.7509765625
                },
                {
                    "start": 75.44,
                    "end": 75.58,
                    "word": " in",
                    "probability": 0.86474609375
                },
                {
                    "start": 75.7,
                    "end": 76.0,
                    "word": " a supporting",
                    "probability": 0.79736328125
                }
            ]
        },
        {
            "id": 13,
            "seek": 7918,
            "start": 76.0,
            "end": 80.18,
            "text": " role here. I'll mostly just be setting the stage for the original work that Young-Hoo is gonna",
            "tokens": [
                50996,
                3090,
                510,
                13,
                286,
                603,
                5240,
                445,
                312,
                3287,
                264,
                3233,
                337,
                264,
                3380,
                589,
                300,
                8160,
                12,
                39,
                1986,
                307,
                799,
                51204
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22120786784739976,
            "compression_ratio": 1.5851528384279476,
            "no_speech_prob": 0.0003299713134765625,
            "words": [
                {
                    "start": 76.0,
                    "end": 76.48,
                    "word": " role",
                    "probability": 0.80126953125
                },
                {
                    "start": 76.48,
                    "end": 76.68,
                    "word": " here",
                    "probability": 0.80419921875
                },
                {
                    "start": 76.68,
                    "end": 76.78,
                    "word": ".",
                    "probability": 0.60009765625
                },
                {
                    "start": 76.78,
                    "end": 76.86,
                    "word": " I'll",
                    "probability": 0.968994140625
                },
                {
                    "start": 76.86,
                    "end": 77.1,
                    "word": " mostly",
                    "probability": 0.78857421875
                },
                {
                    "start": 77.1,
                    "end": 77.38,
                    "word": " just",
                    "probability": 0.86767578125
                },
                {
                    "start": 77.38,
                    "end": 77.6,
                    "word": " be",
                    "probability": 0.91015625
                },
                {
                    "start": 77.6,
                    "end": 77.84,
                    "word": " setting",
                    "probability": 0.85107421875
                },
                {
                    "start": 77.84,
                    "end": 78.02,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 78.02,
                    "end": 78.28,
                    "word": " stage",
                    "probability": 0.72509765625
                },
                {
                    "start": 78.28,
                    "end": 78.58,
                    "word": " for",
                    "probability": 0.89404296875
                },
                {
                    "start": 78.58,
                    "end": 78.86,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 78.86,
                    "end": 79.2,
                    "word": " original",
                    "probability": 0.830078125
                },
                {
                    "start": 79.2,
                    "end": 79.5,
                    "word": " work",
                    "probability": 0.8046875
                },
                {
                    "start": 79.5,
                    "end": 79.66,
                    "word": " that",
                    "probability": 0.87353515625
                },
                {
                    "start": 79.66,
                    "end": 79.82,
                    "word": " Young",
                    "probability": 0.33447265625
                },
                {
                    "start": 79.82,
                    "end": 79.9,
                    "word": "-Hoo",
                    "probability": 0.5987141927083334
                },
                {
                    "start": 79.9,
                    "end": 80.02,
                    "word": " is",
                    "probability": 0.703125
                },
                {
                    "start": 80.02,
                    "end": 80.18,
                    "word": " gonna",
                    "probability": 0.49267578125
                }
            ]
        },
        {
            "id": 14,
            "seek": 7918,
            "start": 80.18,
            "end": 90.04,
            "text": " describe. So, the problem we're solving. Most generally, we're working on the problem of",
            "tokens": [
                51204,
                6786,
                13,
                407,
                11,
                264,
                1154,
                321,
                434,
                12606,
                13,
                4534,
                5101,
                11,
                321,
                434,
                1364,
                322,
                264,
                1154,
                295,
                51584
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22120786784739976,
            "compression_ratio": 1.5851528384279476,
            "no_speech_prob": 0.0003299713134765625,
            "words": [
                {
                    "start": 80.18,
                    "end": 80.76,
                    "word": " describe",
                    "probability": 0.72412109375
                },
                {
                    "start": 80.76,
                    "end": 81.32,
                    "word": ".",
                    "probability": 0.994140625
                },
                {
                    "start": 84.38,
                    "end": 84.92,
                    "word": " So",
                    "probability": 0.869140625
                },
                {
                    "start": 84.92,
                    "end": 85.12,
                    "word": ",",
                    "probability": 0.389404296875
                },
                {
                    "start": 85.12,
                    "end": 85.26,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 85.26,
                    "end": 85.5,
                    "word": " problem",
                    "probability": 0.76513671875
                },
                {
                    "start": 85.5,
                    "end": 85.7,
                    "word": " we're",
                    "probability": 0.9306640625
                },
                {
                    "start": 85.7,
                    "end": 85.98,
                    "word": " solving",
                    "probability": 0.8359375
                },
                {
                    "start": 85.98,
                    "end": 86.54,
                    "word": ".",
                    "probability": 0.87353515625
                },
                {
                    "start": 86.66,
                    "end": 87.2,
                    "word": " Most",
                    "probability": 0.8212890625
                },
                {
                    "start": 87.2,
                    "end": 87.58,
                    "word": " generally",
                    "probability": 0.83349609375
                },
                {
                    "start": 87.58,
                    "end": 87.96,
                    "word": ",",
                    "probability": 0.79052734375
                },
                {
                    "start": 87.96,
                    "end": 88.62,
                    "word": " we're",
                    "probability": 0.923828125
                },
                {
                    "start": 88.62,
                    "end": 89.02,
                    "word": " working",
                    "probability": 0.7587890625
                },
                {
                    "start": 89.02,
                    "end": 89.34,
                    "word": " on",
                    "probability": 0.90625
                },
                {
                    "start": 89.34,
                    "end": 89.58,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 89.58,
                    "end": 89.86,
                    "word": " problem",
                    "probability": 0.7705078125
                },
                {
                    "start": 89.86,
                    "end": 90.04,
                    "word": " of",
                    "probability": 0.9033203125
                }
            ]
        },
        {
            "id": 15,
            "seek": 10918,
            "start": 90.04,
            "end": 96.52,
            "text": " detecting phishing emails. More specifically, we're working on detecting business email",
            "tokens": [
                50364,
                40237,
                903,
                3807,
                12524,
                13,
                5048,
                4682,
                11,
                321,
                434,
                1364,
                322,
                40237,
                1606,
                3796,
                50692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2087928917185933,
            "compression_ratio": 1.8333333333333333,
            "no_speech_prob": 0.04949951171875,
            "words": [
                {
                    "start": 90.04,
                    "end": 90.42,
                    "word": " detecting",
                    "probability": 0.412841796875
                },
                {
                    "start": 90.42,
                    "end": 91.08,
                    "word": " phishing",
                    "probability": 0.88623046875
                },
                {
                    "start": 91.08,
                    "end": 91.42,
                    "word": " emails",
                    "probability": 0.751953125
                },
                {
                    "start": 91.42,
                    "end": 91.94,
                    "word": ".",
                    "probability": 0.8125
                },
                {
                    "start": 91.94,
                    "end": 92.18,
                    "word": " More",
                    "probability": 0.8056640625
                },
                {
                    "start": 92.18,
                    "end": 92.7,
                    "word": " specifically",
                    "probability": 0.72314453125
                },
                {
                    "start": 92.7,
                    "end": 93.18,
                    "word": ",",
                    "probability": 0.77880859375
                },
                {
                    "start": 93.18,
                    "end": 93.46,
                    "word": " we're",
                    "probability": 0.89111328125
                },
                {
                    "start": 93.46,
                    "end": 94.72,
                    "word": " working",
                    "probability": 0.55029296875
                },
                {
                    "start": 94.72,
                    "end": 95.06,
                    "word": " on",
                    "probability": 0.90576171875
                },
                {
                    "start": 95.06,
                    "end": 95.66,
                    "word": " detecting",
                    "probability": 0.78857421875
                },
                {
                    "start": 95.66,
                    "end": 96.28,
                    "word": " business",
                    "probability": 0.8779296875
                },
                {
                    "start": 96.28,
                    "end": 96.52,
                    "word": " email",
                    "probability": 0.8818359375
                }
            ]
        },
        {
            "id": 16,
            "seek": 10918,
            "start": 96.52,
            "end": 102.38,
            "text": " compromise phishing emails and targeted phishing emails. So, to understand what this is, it's",
            "tokens": [
                50692,
                18577,
                903,
                3807,
                12524,
                293,
                15045,
                903,
                3807,
                12524,
                13,
                407,
                11,
                281,
                1223,
                437,
                341,
                307,
                11,
                309,
                311,
                50980
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2087928917185933,
            "compression_ratio": 1.8333333333333333,
            "no_speech_prob": 0.04949951171875,
            "words": [
                {
                    "start": 96.52,
                    "end": 97.14,
                    "word": " compromise",
                    "probability": 0.56591796875
                },
                {
                    "start": 97.14,
                    "end": 97.66,
                    "word": " phishing",
                    "probability": 0.85107421875
                },
                {
                    "start": 97.66,
                    "end": 97.98,
                    "word": " emails",
                    "probability": 0.7333984375
                },
                {
                    "start": 97.98,
                    "end": 98.24,
                    "word": " and",
                    "probability": 0.85400390625
                },
                {
                    "start": 98.24,
                    "end": 98.76,
                    "word": " targeted",
                    "probability": 0.7900390625
                },
                {
                    "start": 98.76,
                    "end": 99.18,
                    "word": " phishing",
                    "probability": 0.9296875
                },
                {
                    "start": 99.18,
                    "end": 99.72,
                    "word": " emails",
                    "probability": 0.7802734375
                },
                {
                    "start": 99.72,
                    "end": 100.34,
                    "word": ".",
                    "probability": 0.9716796875
                },
                {
                    "start": 100.6,
                    "end": 100.68,
                    "word": " So",
                    "probability": 0.75439453125
                },
                {
                    "start": 100.68,
                    "end": 100.82,
                    "word": ",",
                    "probability": 0.1533203125
                },
                {
                    "start": 100.82,
                    "end": 100.86,
                    "word": " to",
                    "probability": 0.8798828125
                },
                {
                    "start": 100.86,
                    "end": 101.22,
                    "word": " understand",
                    "probability": 0.76025390625
                },
                {
                    "start": 101.22,
                    "end": 101.64,
                    "word": " what",
                    "probability": 0.88427734375
                },
                {
                    "start": 101.64,
                    "end": 101.86,
                    "word": " this",
                    "probability": 0.88134765625
                },
                {
                    "start": 101.86,
                    "end": 102.1,
                    "word": " is",
                    "probability": 0.90283203125
                },
                {
                    "start": 102.1,
                    "end": 102.16,
                    "word": ",",
                    "probability": 0.884765625
                },
                {
                    "start": 102.16,
                    "end": 102.38,
                    "word": " it's",
                    "probability": 0.947021484375
                }
            ]
        },
        {
            "id": 17,
            "seek": 10918,
            "start": 102.38,
            "end": 109.1,
            "text": " useful to say what it's not. We're not focused on detecting mass campaigns where detection",
            "tokens": [
                50980,
                4420,
                281,
                584,
                437,
                309,
                311,
                406,
                13,
                492,
                434,
                406,
                5178,
                322,
                40237,
                2758,
                16840,
                689,
                17784,
                51336
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2087928917185933,
            "compression_ratio": 1.8333333333333333,
            "no_speech_prob": 0.04949951171875,
            "words": [
                {
                    "start": 102.38,
                    "end": 102.62,
                    "word": " useful",
                    "probability": 0.73828125
                },
                {
                    "start": 102.62,
                    "end": 102.8,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 102.8,
                    "end": 102.92,
                    "word": " say",
                    "probability": 0.88671875
                },
                {
                    "start": 102.92,
                    "end": 103.06,
                    "word": " what",
                    "probability": 0.87646484375
                },
                {
                    "start": 103.06,
                    "end": 103.26,
                    "word": " it's",
                    "probability": 0.947998046875
                },
                {
                    "start": 103.26,
                    "end": 103.5,
                    "word": " not",
                    "probability": 0.89892578125
                },
                {
                    "start": 103.5,
                    "end": 103.62,
                    "word": ".",
                    "probability": 0.85693359375
                },
                {
                    "start": 103.62,
                    "end": 103.8,
                    "word": " We're",
                    "probability": 0.92529296875
                },
                {
                    "start": 103.8,
                    "end": 103.98,
                    "word": " not",
                    "probability": 0.900390625
                },
                {
                    "start": 103.98,
                    "end": 104.4,
                    "word": " focused",
                    "probability": 0.77099609375
                },
                {
                    "start": 104.4,
                    "end": 105.08,
                    "word": " on",
                    "probability": 0.90625
                },
                {
                    "start": 105.08,
                    "end": 106.0,
                    "word": " detecting",
                    "probability": 0.83154296875
                },
                {
                    "start": 106.0,
                    "end": 106.44,
                    "word": " mass",
                    "probability": 0.798828125
                },
                {
                    "start": 106.44,
                    "end": 107.02,
                    "word": " campaigns",
                    "probability": 0.76171875
                },
                {
                    "start": 107.02,
                    "end": 108.08,
                    "word": " where",
                    "probability": 0.77783203125
                },
                {
                    "start": 108.08,
                    "end": 109.1,
                    "word": " detection",
                    "probability": 0.78955078125
                }
            ]
        },
        {
            "id": 18,
            "seek": 10918,
            "start": 109.1,
            "end": 114.34,
            "text": " reduces to a near-duplicate detection problem because attackers are just sending out millions",
            "tokens": [
                51336,
                18081,
                281,
                257,
                2651,
                12,
                769,
                4770,
                473,
                17784,
                1154,
                570,
                45129,
                366,
                445,
                7750,
                484,
                6803,
                51588
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2087928917185933,
            "compression_ratio": 1.8333333333333333,
            "no_speech_prob": 0.04949951171875,
            "words": [
                {
                    "start": 109.1,
                    "end": 109.9,
                    "word": " reduces",
                    "probability": 0.86767578125
                },
                {
                    "start": 109.9,
                    "end": 110.28,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 110.46,
                    "end": 110.58,
                    "word": " a near",
                    "probability": 0.78271484375
                },
                {
                    "start": 110.58,
                    "end": 111.04,
                    "word": "-duplicate",
                    "probability": 0.82135009765625
                },
                {
                    "start": 111.04,
                    "end": 111.38,
                    "word": " detection",
                    "probability": 0.77734375
                },
                {
                    "start": 111.38,
                    "end": 111.82,
                    "word": " problem",
                    "probability": 0.7822265625
                },
                {
                    "start": 111.82,
                    "end": 112.56,
                    "word": " because",
                    "probability": 0.5703125
                },
                {
                    "start": 112.56,
                    "end": 113.18,
                    "word": " attackers",
                    "probability": 0.58056640625
                },
                {
                    "start": 113.18,
                    "end": 113.4,
                    "word": " are",
                    "probability": 0.89794921875
                },
                {
                    "start": 113.4,
                    "end": 113.54,
                    "word": " just",
                    "probability": 0.8798828125
                },
                {
                    "start": 113.54,
                    "end": 113.78,
                    "word": " sending",
                    "probability": 0.7421875
                },
                {
                    "start": 113.78,
                    "end": 114.0,
                    "word": " out",
                    "probability": 0.80908203125
                },
                {
                    "start": 114.0,
                    "end": 114.34,
                    "word": " millions",
                    "probability": 0.8154296875
                }
            ]
        },
        {
            "id": 19,
            "seek": 10918,
            "start": 114.34,
            "end": 120.02,
            "text": " of copies of basically the same phishing email. Detecting those types of emails turns out to be",
            "tokens": [
                51588,
                295,
                14341,
                295,
                1936,
                264,
                912,
                903,
                3807,
                3796,
                13,
                4237,
                557,
                278,
                729,
                3467,
                295,
                12524,
                4523,
                484,
                281,
                312,
                51856
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2087928917185933,
            "compression_ratio": 1.8333333333333333,
            "no_speech_prob": 0.04949951171875,
            "words": [
                {
                    "start": 114.34,
                    "end": 114.72,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 114.72,
                    "end": 115.02,
                    "word": " copies",
                    "probability": 0.9326171875
                },
                {
                    "start": 115.02,
                    "end": 115.28,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 115.28,
                    "end": 115.66,
                    "word": " basically",
                    "probability": 0.83642578125
                },
                {
                    "start": 115.66,
                    "end": 115.88,
                    "word": " the",
                    "probability": 0.8095703125
                },
                {
                    "start": 115.88,
                    "end": 116.08,
                    "word": " same",
                    "probability": 0.7890625
                },
                {
                    "start": 116.08,
                    "end": 116.36,
                    "word": " phishing",
                    "probability": 0.91162109375
                },
                {
                    "start": 116.36,
                    "end": 116.66,
                    "word": " email",
                    "probability": 0.916015625
                },
                {
                    "start": 116.66,
                    "end": 117.28,
                    "word": ".",
                    "probability": 0.98486328125
                },
                {
                    "start": 117.48,
                    "end": 118.1,
                    "word": " Detecting",
                    "probability": 0.93603515625
                },
                {
                    "start": 118.1,
                    "end": 118.3,
                    "word": " those",
                    "probability": 0.85546875
                },
                {
                    "start": 118.3,
                    "end": 118.8,
                    "word": " types",
                    "probability": 0.7333984375
                },
                {
                    "start": 118.8,
                    "end": 118.94,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 118.94,
                    "end": 119.18,
                    "word": " emails",
                    "probability": 0.771484375
                },
                {
                    "start": 119.18,
                    "end": 119.46,
                    "word": " turns",
                    "probability": 0.6826171875
                },
                {
                    "start": 119.46,
                    "end": 119.62,
                    "word": " out",
                    "probability": 0.8154296875
                },
                {
                    "start": 119.62,
                    "end": 119.74,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 119.74,
                    "end": 120.02,
                    "word": " be",
                    "probability": 0.89794921875
                }
            ]
        },
        {
            "id": 20,
            "seek": 13622,
            "start": 120.2,
            "end": 127.38,
            "text": " a side effect of our focus. But our real focus is on detecting new, custom-authored, bespoke",
            "tokens": [
                50364,
                257,
                1252,
                1802,
                295,
                527,
                1879,
                13,
                583,
                527,
                957,
                1879,
                307,
                322,
                40237,
                777,
                11,
                2375,
                12,
                40198,
                2769,
                11,
                4097,
                48776,
                50752
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24999999732113956,
            "compression_ratio": 1.578512396694215,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 120.2,
                    "end": 120.44,
                    "word": " a side",
                    "probability": 0.79541015625
                },
                {
                    "start": 120.44,
                    "end": 120.76,
                    "word": " effect",
                    "probability": 0.7607421875
                },
                {
                    "start": 120.76,
                    "end": 121.2,
                    "word": " of",
                    "probability": 0.89599609375
                },
                {
                    "start": 121.2,
                    "end": 121.48,
                    "word": " our",
                    "probability": 0.80908203125
                },
                {
                    "start": 121.48,
                    "end": 121.8,
                    "word": " focus",
                    "probability": 0.8359375
                },
                {
                    "start": 121.8,
                    "end": 121.94,
                    "word": ".",
                    "probability": 0.2239990234375
                },
                {
                    "start": 121.94,
                    "end": 122.0,
                    "word": " But",
                    "probability": 0.8369140625
                },
                {
                    "start": 122.0,
                    "end": 122.22,
                    "word": " our",
                    "probability": 0.7734375
                },
                {
                    "start": 122.22,
                    "end": 122.4,
                    "word": " real",
                    "probability": 0.88232421875
                },
                {
                    "start": 122.4,
                    "end": 122.72,
                    "word": " focus",
                    "probability": 0.833984375
                },
                {
                    "start": 122.72,
                    "end": 122.94,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 122.94,
                    "end": 123.16,
                    "word": " on",
                    "probability": 0.90576171875
                },
                {
                    "start": 123.16,
                    "end": 124.02,
                    "word": " detecting",
                    "probability": 0.7880859375
                },
                {
                    "start": 124.02,
                    "end": 124.68,
                    "word": " new",
                    "probability": 0.83740234375
                },
                {
                    "start": 124.68,
                    "end": 124.88,
                    "word": ",",
                    "probability": 0.1968994140625
                },
                {
                    "start": 124.88,
                    "end": 125.2,
                    "word": " custom",
                    "probability": 0.75244140625
                },
                {
                    "start": 125.2,
                    "end": 125.74,
                    "word": "-authored",
                    "probability": 0.7413736979166666
                },
                {
                    "start": 125.74,
                    "end": 126.18,
                    "word": ",",
                    "probability": 0.8193359375
                },
                {
                    "start": 126.18,
                    "end": 127.38,
                    "word": " bespoke",
                    "probability": 0.843994140625
                }
            ]
        },
        {
            "id": 21,
            "seek": 13622,
            "start": 127.38,
            "end": 133.2,
            "text": " phishing emails that are based on research on a target. And this diagram does a good job of",
            "tokens": [
                50752,
                903,
                3807,
                12524,
                300,
                366,
                2361,
                322,
                2132,
                322,
                257,
                3779,
                13,
                400,
                341,
                10686,
                775,
                257,
                665,
                1691,
                295,
                51016
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24999999732113956,
            "compression_ratio": 1.578512396694215,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 127.38,
                    "end": 128.16,
                    "word": " phishing",
                    "probability": 0.903564453125
                },
                {
                    "start": 128.16,
                    "end": 128.48,
                    "word": " emails",
                    "probability": 0.7294921875
                },
                {
                    "start": 128.48,
                    "end": 128.9,
                    "word": " that",
                    "probability": 0.8544921875
                },
                {
                    "start": 128.9,
                    "end": 129.0,
                    "word": " are",
                    "probability": 0.89208984375
                },
                {
                    "start": 129.0,
                    "end": 129.2,
                    "word": " based",
                    "probability": 0.82421875
                },
                {
                    "start": 129.2,
                    "end": 129.38,
                    "word": " on",
                    "probability": 0.9140625
                },
                {
                    "start": 129.38,
                    "end": 129.8,
                    "word": " research",
                    "probability": 0.751953125
                },
                {
                    "start": 129.8,
                    "end": 130.1,
                    "word": " on",
                    "probability": 0.90185546875
                },
                {
                    "start": 130.22,
                    "end": 130.46,
                    "word": " a target",
                    "probability": 0.8408203125
                },
                {
                    "start": 130.46,
                    "end": 131.1,
                    "word": ".",
                    "probability": 0.95556640625
                },
                {
                    "start": 131.28,
                    "end": 131.62,
                    "word": " And",
                    "probability": 0.7646484375
                },
                {
                    "start": 131.62,
                    "end": 132.14,
                    "word": " this",
                    "probability": 0.8681640625
                },
                {
                    "start": 132.14,
                    "end": 132.5,
                    "word": " diagram",
                    "probability": 0.74853515625
                },
                {
                    "start": 132.5,
                    "end": 132.68,
                    "word": " does",
                    "probability": 0.919921875
                },
                {
                    "start": 132.78,
                    "end": 132.88,
                    "word": " a good",
                    "probability": 0.88134765625
                },
                {
                    "start": 132.88,
                    "end": 133.08,
                    "word": " job",
                    "probability": 0.90771484375
                },
                {
                    "start": 133.08,
                    "end": 133.2,
                    "word": " of",
                    "probability": 0.89501953125
                }
            ]
        },
        {
            "id": 22,
            "seek": 13622,
            "start": 133.2,
            "end": 138.84,
            "text": " getting across the kind of workflow we're looking at stopping. In this workflow, cyber criminals and",
            "tokens": [
                51016,
                1242,
                2108,
                264,
                733,
                295,
                20993,
                321,
                434,
                1237,
                412,
                12767,
                13,
                682,
                341,
                20993,
                11,
                13411,
                23474,
                293,
                51300
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24999999732113956,
            "compression_ratio": 1.578512396694215,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 133.2,
                    "end": 133.36,
                    "word": " getting",
                    "probability": 0.90234375
                },
                {
                    "start": 133.36,
                    "end": 133.78,
                    "word": " across",
                    "probability": 0.7041015625
                },
                {
                    "start": 133.78,
                    "end": 133.98,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 133.98,
                    "end": 134.12,
                    "word": " kind",
                    "probability": 0.72265625
                },
                {
                    "start": 134.12,
                    "end": 134.22,
                    "word": " of",
                    "probability": 0.89794921875
                },
                {
                    "start": 134.22,
                    "end": 134.52,
                    "word": " workflow",
                    "probability": 0.74267578125
                },
                {
                    "start": 134.52,
                    "end": 134.88,
                    "word": " we're",
                    "probability": 0.8134765625
                },
                {
                    "start": 134.88,
                    "end": 135.1,
                    "word": " looking",
                    "probability": 0.78515625
                },
                {
                    "start": 135.1,
                    "end": 135.24,
                    "word": " at",
                    "probability": 0.90087890625
                },
                {
                    "start": 135.24,
                    "end": 135.58,
                    "word": " stopping",
                    "probability": 0.84423828125
                },
                {
                    "start": 135.58,
                    "end": 136.22,
                    "word": ".",
                    "probability": 0.9365234375
                },
                {
                    "start": 136.38,
                    "end": 136.54,
                    "word": " In",
                    "probability": 0.56201171875
                },
                {
                    "start": 136.54,
                    "end": 136.74,
                    "word": " this",
                    "probability": 0.88037109375
                },
                {
                    "start": 136.74,
                    "end": 137.16,
                    "word": " workflow",
                    "probability": 0.71923828125
                },
                {
                    "start": 137.16,
                    "end": 137.72,
                    "word": ",",
                    "probability": 0.9482421875
                },
                {
                    "start": 137.72,
                    "end": 138.24,
                    "word": " cyber",
                    "probability": 0.9169921875
                },
                {
                    "start": 138.24,
                    "end": 138.62,
                    "word": " criminals",
                    "probability": 0.415283203125
                },
                {
                    "start": 138.62,
                    "end": 138.84,
                    "word": " and",
                    "probability": 0.888671875
                }
            ]
        },
        {
            "id": 23,
            "seek": 13622,
            "start": 138.84,
            "end": 147.08,
            "text": " attackers identify targets through open-source research, usually on the web. Then they establish",
            "tokens": [
                51300,
                45129,
                5876,
                12911,
                807,
                1269,
                12,
                41676,
                2132,
                11,
                2673,
                322,
                264,
                3670,
                13,
                1396,
                436,
                8327,
                51720
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24999999732113956,
            "compression_ratio": 1.578512396694215,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 138.84,
                    "end": 139.2,
                    "word": " attackers",
                    "probability": 0.82470703125
                },
                {
                    "start": 139.2,
                    "end": 140.34,
                    "word": " identify",
                    "probability": 0.876953125
                },
                {
                    "start": 140.34,
                    "end": 140.92,
                    "word": " targets",
                    "probability": 0.68603515625
                },
                {
                    "start": 140.92,
                    "end": 141.46,
                    "word": " through",
                    "probability": 0.78662109375
                },
                {
                    "start": 141.46,
                    "end": 141.8,
                    "word": " open",
                    "probability": 0.78173828125
                },
                {
                    "start": 141.8,
                    "end": 142.02,
                    "word": "-source",
                    "probability": 0.615966796875
                },
                {
                    "start": 142.02,
                    "end": 142.56,
                    "word": " research",
                    "probability": 0.77880859375
                },
                {
                    "start": 142.56,
                    "end": 143.2,
                    "word": ",",
                    "probability": 0.81982421875
                },
                {
                    "start": 143.2,
                    "end": 143.94,
                    "word": " usually",
                    "probability": 0.83154296875
                },
                {
                    "start": 143.94,
                    "end": 144.38,
                    "word": " on",
                    "probability": 0.89892578125
                },
                {
                    "start": 144.38,
                    "end": 144.52,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 144.52,
                    "end": 144.76,
                    "word": " web",
                    "probability": 0.71435546875
                },
                {
                    "start": 144.76,
                    "end": 145.4,
                    "word": ".",
                    "probability": 0.75390625
                },
                {
                    "start": 145.56,
                    "end": 145.9,
                    "word": " Then",
                    "probability": 0.8125
                },
                {
                    "start": 145.9,
                    "end": 146.78,
                    "word": " they",
                    "probability": 0.71142578125
                },
                {
                    "start": 146.78,
                    "end": 147.08,
                    "word": " establish",
                    "probability": 0.80908203125
                }
            ]
        },
        {
            "id": 24,
            "seek": 16400,
            "start": 147.08,
            "end": 154.98,
            "text": " contact with those targets. In step two, we call this grooming. In this case, they put out a lure,",
            "tokens": [
                50368,
                3385,
                365,
                729,
                12911,
                13,
                682,
                1823,
                732,
                11,
                321,
                818,
                341,
                49700,
                13,
                682,
                341,
                1389,
                11,
                436,
                829,
                484,
                257,
                32350,
                11,
                50752
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1993475215775626,
            "compression_ratio": 1.7935779816513762,
            "no_speech_prob": 0.00856781005859375,
            "words": [
                {
                    "start": 147.08,
                    "end": 147.44,
                    "word": " contact",
                    "probability": 0.06207275390625
                },
                {
                    "start": 147.44,
                    "end": 148.08,
                    "word": " with",
                    "probability": 0.8193359375
                },
                {
                    "start": 148.08,
                    "end": 148.56,
                    "word": " those",
                    "probability": 0.5302734375
                },
                {
                    "start": 148.56,
                    "end": 149.06,
                    "word": " targets",
                    "probability": 0.63134765625
                },
                {
                    "start": 149.06,
                    "end": 149.62,
                    "word": ".",
                    "probability": 0.8232421875
                },
                {
                    "start": 149.98,
                    "end": 150.24,
                    "word": " In",
                    "probability": 0.81982421875
                },
                {
                    "start": 150.24,
                    "end": 150.42,
                    "word": " step",
                    "probability": 0.6435546875
                },
                {
                    "start": 150.42,
                    "end": 150.66,
                    "word": " two",
                    "probability": 0.61279296875
                },
                {
                    "start": 150.66,
                    "end": 150.92,
                    "word": ",",
                    "probability": 0.64453125
                },
                {
                    "start": 150.92,
                    "end": 151.3,
                    "word": " we",
                    "probability": 0.8818359375
                },
                {
                    "start": 151.3,
                    "end": 151.44,
                    "word": " call",
                    "probability": 0.794921875
                },
                {
                    "start": 151.44,
                    "end": 151.62,
                    "word": " this",
                    "probability": 0.869140625
                },
                {
                    "start": 151.62,
                    "end": 151.94,
                    "word": " grooming",
                    "probability": 0.603515625
                },
                {
                    "start": 151.94,
                    "end": 152.5,
                    "word": ".",
                    "probability": 0.771484375
                },
                {
                    "start": 152.94,
                    "end": 153.36,
                    "word": " In",
                    "probability": 0.90771484375
                },
                {
                    "start": 153.36,
                    "end": 153.6,
                    "word": " this",
                    "probability": 0.88037109375
                },
                {
                    "start": 153.6,
                    "end": 153.78,
                    "word": " case",
                    "probability": 0.8046875
                },
                {
                    "start": 153.78,
                    "end": 153.84,
                    "word": ",",
                    "probability": 0.6083984375
                },
                {
                    "start": 153.84,
                    "end": 153.98,
                    "word": " they",
                    "probability": 0.73779296875
                },
                {
                    "start": 153.98,
                    "end": 154.22,
                    "word": " put",
                    "probability": 0.8828125
                },
                {
                    "start": 154.22,
                    "end": 154.36,
                    "word": " out",
                    "probability": 0.76220703125
                },
                {
                    "start": 154.5,
                    "end": 154.7,
                    "word": " a lure",
                    "probability": 0.9169921875
                },
                {
                    "start": 154.7,
                    "end": 154.98,
                    "word": ",",
                    "probability": 0.77392578125
                }
            ]
        },
        {
            "id": 25,
            "seek": 16400,
            "start": 154.98,
            "end": 161.28,
            "text": " an initial email, usually or sometimes an initial text message, and then build trust and authenticity",
            "tokens": [
                50752,
                364,
                5883,
                3796,
                11,
                2673,
                420,
                2171,
                364,
                5883,
                2487,
                3636,
                11,
                293,
                550,
                1322,
                3361,
                293,
                34215,
                51092
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1993475215775626,
            "compression_ratio": 1.7935779816513762,
            "no_speech_prob": 0.00856781005859375,
            "words": [
                {
                    "start": 154.98,
                    "end": 155.14,
                    "word": " an",
                    "probability": 0.86181640625
                },
                {
                    "start": 155.14,
                    "end": 155.42,
                    "word": " initial",
                    "probability": 0.85693359375
                },
                {
                    "start": 155.42,
                    "end": 155.78,
                    "word": " email",
                    "probability": 0.86181640625
                },
                {
                    "start": 155.78,
                    "end": 156.02,
                    "word": ",",
                    "probability": 0.66650390625
                },
                {
                    "start": 156.02,
                    "end": 156.34,
                    "word": " usually",
                    "probability": 0.82861328125
                },
                {
                    "start": 156.34,
                    "end": 156.52,
                    "word": " or",
                    "probability": 0.7216796875
                },
                {
                    "start": 156.52,
                    "end": 156.82,
                    "word": " sometimes",
                    "probability": 0.88330078125
                },
                {
                    "start": 156.82,
                    "end": 157.02,
                    "word": " an",
                    "probability": 0.80908203125
                },
                {
                    "start": 157.02,
                    "end": 157.28,
                    "word": " initial",
                    "probability": 0.880859375
                },
                {
                    "start": 157.28,
                    "end": 157.56,
                    "word": " text",
                    "probability": 0.90380859375
                },
                {
                    "start": 157.56,
                    "end": 158.0,
                    "word": " message",
                    "probability": 0.82861328125
                },
                {
                    "start": 158.0,
                    "end": 158.56,
                    "word": ",",
                    "probability": 0.724609375
                },
                {
                    "start": 158.56,
                    "end": 158.88,
                    "word": " and",
                    "probability": 0.8876953125
                },
                {
                    "start": 158.88,
                    "end": 159.04,
                    "word": " then",
                    "probability": 0.73388671875
                },
                {
                    "start": 159.04,
                    "end": 159.28,
                    "word": " build",
                    "probability": 0.8740234375
                },
                {
                    "start": 159.28,
                    "end": 159.7,
                    "word": " trust",
                    "probability": 0.8046875
                },
                {
                    "start": 159.7,
                    "end": 160.36,
                    "word": " and",
                    "probability": 0.888671875
                },
                {
                    "start": 160.36,
                    "end": 161.28,
                    "word": " authenticity",
                    "probability": 0.552734375
                }
            ]
        },
        {
            "id": 26,
            "seek": 16400,
            "start": 161.28,
            "end": 169.5,
            "text": " around some identity that they're impersonating with the mark. In step three, they cash out the",
            "tokens": [
                51092,
                926,
                512,
                6575,
                300,
                436,
                434,
                38147,
                990,
                365,
                264,
                1491,
                13,
                682,
                1823,
                1045,
                11,
                436,
                6388,
                484,
                264,
                51480
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1993475215775626,
            "compression_ratio": 1.7935779816513762,
            "no_speech_prob": 0.00856781005859375,
            "words": [
                {
                    "start": 161.28,
                    "end": 162.02,
                    "word": " around",
                    "probability": 0.85888671875
                },
                {
                    "start": 162.02,
                    "end": 162.4,
                    "word": " some",
                    "probability": 0.80859375
                },
                {
                    "start": 162.4,
                    "end": 162.74,
                    "word": " identity",
                    "probability": 0.84814453125
                },
                {
                    "start": 162.74,
                    "end": 162.96,
                    "word": " that",
                    "probability": 0.794921875
                },
                {
                    "start": 162.96,
                    "end": 163.12,
                    "word": " they're",
                    "probability": 0.818359375
                },
                {
                    "start": 163.12,
                    "end": 163.72,
                    "word": " impersonating",
                    "probability": 0.89697265625
                },
                {
                    "start": 163.72,
                    "end": 164.4,
                    "word": " with",
                    "probability": 0.7705078125
                },
                {
                    "start": 164.4,
                    "end": 164.8,
                    "word": " the",
                    "probability": 0.693359375
                },
                {
                    "start": 164.8,
                    "end": 165.06,
                    "word": " mark",
                    "probability": 0.80322265625
                },
                {
                    "start": 165.06,
                    "end": 165.62,
                    "word": ".",
                    "probability": 0.99462890625
                },
                {
                    "start": 165.8,
                    "end": 166.02,
                    "word": " In",
                    "probability": 0.912109375
                },
                {
                    "start": 166.02,
                    "end": 166.22,
                    "word": " step",
                    "probability": 0.81787109375
                },
                {
                    "start": 166.22,
                    "end": 166.54,
                    "word": " three",
                    "probability": 0.8330078125
                },
                {
                    "start": 166.54,
                    "end": 167.0,
                    "word": ",",
                    "probability": 0.955078125
                },
                {
                    "start": 167.0,
                    "end": 167.6,
                    "word": " they",
                    "probability": 0.73486328125
                },
                {
                    "start": 167.6,
                    "end": 168.92,
                    "word": " cash",
                    "probability": 0.7978515625
                },
                {
                    "start": 168.92,
                    "end": 169.2,
                    "word": " out",
                    "probability": 0.796875
                },
                {
                    "start": 169.2,
                    "end": 169.5,
                    "word": " the",
                    "probability": 0.8125
                }
            ]
        },
        {
            "id": 27,
            "seek": 16400,
            "start": 169.5,
            "end": 174.86,
            "text": " trust that they've built with their targets and make an ask of those targets. Oftentimes, that",
            "tokens": [
                51480,
                3361,
                300,
                436,
                600,
                3094,
                365,
                641,
                12911,
                293,
                652,
                364,
                1029,
                295,
                729,
                12911,
                13,
                46636,
                11,
                300,
                51748
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1993475215775626,
            "compression_ratio": 1.7935779816513762,
            "no_speech_prob": 0.00856781005859375,
            "words": [
                {
                    "start": 169.5,
                    "end": 169.8,
                    "word": " trust",
                    "probability": 0.81787109375
                },
                {
                    "start": 169.8,
                    "end": 169.92,
                    "word": " that",
                    "probability": 0.88037109375
                },
                {
                    "start": 169.92,
                    "end": 170.14,
                    "word": " they've",
                    "probability": 0.798095703125
                },
                {
                    "start": 170.14,
                    "end": 170.28,
                    "word": " built",
                    "probability": 0.927734375
                },
                {
                    "start": 170.28,
                    "end": 170.46,
                    "word": " with",
                    "probability": 0.81689453125
                },
                {
                    "start": 170.46,
                    "end": 170.6,
                    "word": " their",
                    "probability": 0.904296875
                },
                {
                    "start": 170.6,
                    "end": 171.04,
                    "word": " targets",
                    "probability": 0.66943359375
                },
                {
                    "start": 171.04,
                    "end": 172.06,
                    "word": " and",
                    "probability": 0.765625
                },
                {
                    "start": 172.06,
                    "end": 172.56,
                    "word": " make",
                    "probability": 0.86962890625
                },
                {
                    "start": 172.56,
                    "end": 172.72,
                    "word": " an",
                    "probability": 0.9111328125
                },
                {
                    "start": 172.72,
                    "end": 173.02,
                    "word": " ask",
                    "probability": 0.7939453125
                },
                {
                    "start": 173.02,
                    "end": 173.22,
                    "word": " of",
                    "probability": 0.8857421875
                },
                {
                    "start": 173.22,
                    "end": 173.4,
                    "word": " those",
                    "probability": 0.853515625
                },
                {
                    "start": 173.4,
                    "end": 173.8,
                    "word": " targets",
                    "probability": 0.65869140625
                },
                {
                    "start": 173.8,
                    "end": 174.1,
                    "word": ".",
                    "probability": 0.9130859375
                },
                {
                    "start": 174.1,
                    "end": 174.38,
                    "word": " Oftentimes",
                    "probability": 0.8212890625
                },
                {
                    "start": 174.38,
                    "end": 174.76,
                    "word": ",",
                    "probability": 0.452880859375
                },
                {
                    "start": 174.76,
                    "end": 174.86,
                    "word": " that",
                    "probability": 0.884765625
                }
            ]
        },
        {
            "id": 28,
            "seek": 19236,
            "start": 174.86,
            "end": 179.5,
            "text": " will be around wiring money or sending credentials. And then in step four, they actually",
            "tokens": [
                50384,
                486,
                312,
                926,
                27520,
                1460,
                420,
                7750,
                27404,
                13,
                400,
                550,
                294,
                1823,
                1451,
                11,
                436,
                767,
                50596
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2381465548071368,
            "compression_ratio": 1.5958333333333334,
            "no_speech_prob": 0.0012445449829101562,
            "words": [
                {
                    "start": 174.86,
                    "end": 175.12,
                    "word": " will",
                    "probability": 0.0146636962890625
                },
                {
                    "start": 175.12,
                    "end": 175.66,
                    "word": " be",
                    "probability": 0.865234375
                },
                {
                    "start": 175.66,
                    "end": 176.02,
                    "word": " around",
                    "probability": 0.8671875
                },
                {
                    "start": 176.02,
                    "end": 176.4,
                    "word": " wiring",
                    "probability": 0.69873046875
                },
                {
                    "start": 176.4,
                    "end": 176.78,
                    "word": " money",
                    "probability": 0.8544921875
                },
                {
                    "start": 176.78,
                    "end": 177.22,
                    "word": " or",
                    "probability": 0.814453125
                },
                {
                    "start": 177.22,
                    "end": 177.64,
                    "word": " sending",
                    "probability": 0.70361328125
                },
                {
                    "start": 177.64,
                    "end": 178.06,
                    "word": " credentials",
                    "probability": 0.79296875
                },
                {
                    "start": 178.06,
                    "end": 178.36,
                    "word": ".",
                    "probability": 0.248779296875
                },
                {
                    "start": 178.36,
                    "end": 178.38,
                    "word": " And",
                    "probability": 0.76025390625
                },
                {
                    "start": 178.38,
                    "end": 178.5,
                    "word": " then",
                    "probability": 0.7109375
                },
                {
                    "start": 178.5,
                    "end": 178.62,
                    "word": " in",
                    "probability": 0.08203125
                },
                {
                    "start": 178.62,
                    "end": 178.76,
                    "word": " step",
                    "probability": 0.6328125
                },
                {
                    "start": 178.76,
                    "end": 178.96,
                    "word": " four",
                    "probability": 0.6806640625
                },
                {
                    "start": 178.96,
                    "end": 179.02,
                    "word": ",",
                    "probability": 0.66748046875
                },
                {
                    "start": 179.02,
                    "end": 179.1,
                    "word": " they",
                    "probability": 0.73681640625
                },
                {
                    "start": 179.1,
                    "end": 179.5,
                    "word": " actually",
                    "probability": 0.7744140625
                }
            ]
        },
        {
            "id": 29,
            "seek": 19236,
            "start": 181.82,
            "end": 184.05,
            "text": " receive money or receive the credentials.",
            "tokens": [
                50720,
                4774,
                1460,
                420,
                4774,
                264,
                27404,
                13,
                50812
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2381465548071368,
            "compression_ratio": 1.5958333333333334,
            "no_speech_prob": 0.0012445449829101562,
            "words": [
                {
                    "start": 181.82,
                    "end": 182.4,
                    "word": " receive",
                    "probability": 0.82958984375
                },
                {
                    "start": 182.4,
                    "end": 182.72,
                    "word": " money",
                    "probability": 0.814453125
                },
                {
                    "start": 182.72,
                    "end": 183.0,
                    "word": " or",
                    "probability": 0.8369140625
                },
                {
                    "start": 183.0,
                    "end": 183.26,
                    "word": " receive",
                    "probability": 0.9052734375
                },
                {
                    "start": 183.26,
                    "end": 183.4,
                    "word": " the",
                    "probability": 0.75927734375
                },
                {
                    "start": 183.4,
                    "end": 183.76,
                    "word": " credentials",
                    "probability": 0.7529296875
                },
                {
                    "start": 183.76,
                    "end": 184.05,
                    "word": ".",
                    "probability": 0.99169921875
                }
            ]
        },
        {
            "id": 30,
            "seek": 19236,
            "start": 187.02,
            "end": 191.54,
            "text": " So within targeted phishing, business email compromise, which is focused on",
            "tokens": [
                50968,
                407,
                1951,
                15045,
                903,
                3807,
                11,
                1606,
                3796,
                18577,
                11,
                597,
                307,
                5178,
                322,
                51208
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2381465548071368,
            "compression_ratio": 1.5958333333333334,
            "no_speech_prob": 0.0012445449829101562,
            "words": [
                {
                    "start": 187.02,
                    "end": 187.6,
                    "word": " So",
                    "probability": 0.80712890625
                },
                {
                    "start": 187.6,
                    "end": 188.18,
                    "word": " within",
                    "probability": 0.5888671875
                },
                {
                    "start": 188.18,
                    "end": 188.6,
                    "word": " targeted",
                    "probability": 0.69775390625
                },
                {
                    "start": 188.6,
                    "end": 189.02,
                    "word": " phishing",
                    "probability": 0.878662109375
                },
                {
                    "start": 189.02,
                    "end": 189.22,
                    "word": ",",
                    "probability": 0.8486328125
                },
                {
                    "start": 189.22,
                    "end": 189.58,
                    "word": " business",
                    "probability": 0.88330078125
                },
                {
                    "start": 189.58,
                    "end": 189.82,
                    "word": " email",
                    "probability": 0.87890625
                },
                {
                    "start": 189.82,
                    "end": 190.2,
                    "word": " compromise",
                    "probability": 0.65478515625
                },
                {
                    "start": 190.2,
                    "end": 190.5,
                    "word": ",",
                    "probability": 0.8662109375
                },
                {
                    "start": 190.5,
                    "end": 190.64,
                    "word": " which",
                    "probability": 0.919921875
                },
                {
                    "start": 190.64,
                    "end": 190.8,
                    "word": " is",
                    "probability": 0.87158203125
                },
                {
                    "start": 190.8,
                    "end": 191.06,
                    "word": " focused",
                    "probability": 0.75244140625
                },
                {
                    "start": 191.06,
                    "end": 191.54,
                    "word": " on",
                    "probability": 0.9091796875
                }
            ]
        },
        {
            "id": 31,
            "seek": 19236,
            "start": 191.54,
            "end": 197.16,
            "text": " stealing money from businesses and other organizations, has been a growing trend.",
            "tokens": [
                51208,
                19757,
                1460,
                490,
                6011,
                293,
                661,
                6150,
                11,
                575,
                668,
                257,
                4194,
                6028,
                13,
                51448
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2381465548071368,
            "compression_ratio": 1.5958333333333334,
            "no_speech_prob": 0.0012445449829101562,
            "words": [
                {
                    "start": 191.54,
                    "end": 192.14,
                    "word": " stealing",
                    "probability": 0.736328125
                },
                {
                    "start": 192.14,
                    "end": 192.5,
                    "word": " money",
                    "probability": 0.84228515625
                },
                {
                    "start": 192.5,
                    "end": 193.06,
                    "word": " from",
                    "probability": 0.81494140625
                },
                {
                    "start": 193.06,
                    "end": 193.72,
                    "word": " businesses",
                    "probability": 0.78173828125
                },
                {
                    "start": 193.72,
                    "end": 193.94,
                    "word": " and",
                    "probability": 0.88232421875
                },
                {
                    "start": 193.94,
                    "end": 194.12,
                    "word": " other",
                    "probability": 0.78271484375
                },
                {
                    "start": 194.12,
                    "end": 194.68,
                    "word": " organizations",
                    "probability": 0.69580078125
                },
                {
                    "start": 194.68,
                    "end": 195.26,
                    "word": ",",
                    "probability": 0.7314453125
                },
                {
                    "start": 195.26,
                    "end": 195.44,
                    "word": " has",
                    "probability": 0.8955078125
                },
                {
                    "start": 195.44,
                    "end": 195.7,
                    "word": " been",
                    "probability": 0.82275390625
                },
                {
                    "start": 195.98,
                    "end": 196.18,
                    "word": " a growing",
                    "probability": 0.7099609375
                },
                {
                    "start": 196.18,
                    "end": 196.64,
                    "word": " trend",
                    "probability": 0.74658203125
                },
                {
                    "start": 196.64,
                    "end": 197.16,
                    "word": ".",
                    "probability": 0.99462890625
                }
            ]
        },
        {
            "id": 32,
            "seek": 19236,
            "start": 197.16,
            "end": 203.22,
            "text": " So as you can see here, in July 2016, a few billion dollars were stolen, according to the FBI,",
            "tokens": [
                51484,
                407,
                382,
                291,
                393,
                536,
                510,
                11,
                294,
                7370,
                6549,
                11,
                257,
                1326,
                5218,
                3808,
                645,
                15900,
                11,
                4650,
                281,
                264,
                17441,
                11,
                51752
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2381465548071368,
            "compression_ratio": 1.5958333333333334,
            "no_speech_prob": 0.0012445449829101562,
            "words": [
                {
                    "start": 197.16,
                    "end": 197.58,
                    "word": " So",
                    "probability": 0.76611328125
                },
                {
                    "start": 197.58,
                    "end": 197.8,
                    "word": " as",
                    "probability": 0.81787109375
                },
                {
                    "start": 197.8,
                    "end": 197.9,
                    "word": " you",
                    "probability": 0.8935546875
                },
                {
                    "start": 197.9,
                    "end": 198.02,
                    "word": " can",
                    "probability": 0.8994140625
                },
                {
                    "start": 198.02,
                    "end": 198.18,
                    "word": " see",
                    "probability": 0.81591796875
                },
                {
                    "start": 198.18,
                    "end": 198.38,
                    "word": " here",
                    "probability": 0.810546875
                },
                {
                    "start": 198.38,
                    "end": 198.42,
                    "word": ",",
                    "probability": 0.5654296875
                },
                {
                    "start": 198.42,
                    "end": 198.54,
                    "word": " in",
                    "probability": 0.89794921875
                },
                {
                    "start": 198.54,
                    "end": 198.72,
                    "word": " July",
                    "probability": 0.83837890625
                },
                {
                    "start": 198.72,
                    "end": 199.34,
                    "word": " 2016",
                    "probability": 0.61962890625
                },
                {
                    "start": 199.34,
                    "end": 199.94,
                    "word": ",",
                    "probability": 0.9482421875
                },
                {
                    "start": 200.28,
                    "end": 200.42,
                    "word": " a few",
                    "probability": 0.82080078125
                },
                {
                    "start": 200.42,
                    "end": 200.72,
                    "word": " billion",
                    "probability": 0.78515625
                },
                {
                    "start": 200.72,
                    "end": 201.12,
                    "word": " dollars",
                    "probability": 0.908203125
                },
                {
                    "start": 201.12,
                    "end": 201.36,
                    "word": " were",
                    "probability": 0.875
                },
                {
                    "start": 201.36,
                    "end": 201.64,
                    "word": " stolen",
                    "probability": 0.75341796875
                },
                {
                    "start": 201.64,
                    "end": 201.76,
                    "word": ",",
                    "probability": 0.6845703125
                },
                {
                    "start": 201.76,
                    "end": 201.96,
                    "word": " according",
                    "probability": 0.8408203125
                },
                {
                    "start": 201.96,
                    "end": 202.12,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 202.12,
                    "end": 202.26,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 202.26,
                    "end": 202.7,
                    "word": " FBI",
                    "probability": 0.94384765625
                },
                {
                    "start": 202.7,
                    "end": 203.22,
                    "word": ",",
                    "probability": 0.81298828125
                }
            ]
        },
        {
            "id": 33,
            "seek": 21960,
            "start": 203.22,
            "end": 207.72,
            "text": " through business email compromise attacks. These are targeted phishing attacks that extort money",
            "tokens": [
                50376,
                807,
                1606,
                3796,
                18577,
                8122,
                13,
                1981,
                366,
                15045,
                903,
                3807,
                8122,
                300,
                1279,
                477,
                1460,
                50584
            ],
            "temperature": 0.0,
            "avg_logprob": -0.17851217025149185,
            "compression_ratio": 1.6175438596491227,
            "no_speech_prob": 0.0006260871887207031,
            "words": [
                {
                    "start": 203.22,
                    "end": 203.74,
                    "word": " through",
                    "probability": 0.36474609375
                },
                {
                    "start": 203.74,
                    "end": 204.08,
                    "word": " business",
                    "probability": 0.8603515625
                },
                {
                    "start": 204.08,
                    "end": 204.3,
                    "word": " email",
                    "probability": 0.83544921875
                },
                {
                    "start": 204.3,
                    "end": 204.7,
                    "word": " compromise",
                    "probability": 0.58447265625
                },
                {
                    "start": 204.7,
                    "end": 205.22,
                    "word": " attacks",
                    "probability": 0.72412109375
                },
                {
                    "start": 205.22,
                    "end": 205.52,
                    "word": ".",
                    "probability": 0.68603515625
                },
                {
                    "start": 205.52,
                    "end": 205.74,
                    "word": " These",
                    "probability": 0.703125
                },
                {
                    "start": 205.74,
                    "end": 205.84,
                    "word": " are",
                    "probability": 0.89599609375
                },
                {
                    "start": 205.84,
                    "end": 206.16,
                    "word": " targeted",
                    "probability": 0.79833984375
                },
                {
                    "start": 206.16,
                    "end": 206.48,
                    "word": " phishing",
                    "probability": 0.88525390625
                },
                {
                    "start": 206.48,
                    "end": 206.8,
                    "word": " attacks",
                    "probability": 0.7314453125
                },
                {
                    "start": 206.8,
                    "end": 207.0,
                    "word": " that",
                    "probability": 0.87646484375
                },
                {
                    "start": 207.0,
                    "end": 207.3,
                    "word": " extort",
                    "probability": 0.883544921875
                },
                {
                    "start": 207.3,
                    "end": 207.72,
                    "word": " money",
                    "probability": 0.841796875
                }
            ]
        },
        {
            "id": 34,
            "seek": 21960,
            "start": 207.72,
            "end": 215.02,
            "text": " from organizations. In May 2017, that number had grown to almost six billion. And in 2018,",
            "tokens": [
                50584,
                490,
                6150,
                13,
                682,
                1891,
                6591,
                11,
                300,
                1230,
                632,
                7709,
                281,
                1920,
                2309,
                5218,
                13,
                400,
                294,
                6096,
                11,
                50944
            ],
            "temperature": 0.0,
            "avg_logprob": -0.17851217025149185,
            "compression_ratio": 1.6175438596491227,
            "no_speech_prob": 0.0006260871887207031,
            "words": [
                {
                    "start": 207.72,
                    "end": 207.94,
                    "word": " from",
                    "probability": 0.814453125
                },
                {
                    "start": 207.94,
                    "end": 208.4,
                    "word": " organizations",
                    "probability": 0.685546875
                },
                {
                    "start": 208.4,
                    "end": 208.88,
                    "word": ".",
                    "probability": 0.94140625
                },
                {
                    "start": 209.3,
                    "end": 209.42,
                    "word": " In",
                    "probability": 0.87158203125
                },
                {
                    "start": 209.42,
                    "end": 209.8,
                    "word": " May",
                    "probability": 0.86669921875
                },
                {
                    "start": 209.8,
                    "end": 210.4,
                    "word": " 2017",
                    "probability": 0.6826171875
                },
                {
                    "start": 210.4,
                    "end": 210.76,
                    "word": ",",
                    "probability": 0.69580078125
                },
                {
                    "start": 210.76,
                    "end": 210.94,
                    "word": " that",
                    "probability": 0.88427734375
                },
                {
                    "start": 210.94,
                    "end": 211.2,
                    "word": " number",
                    "probability": 0.8740234375
                },
                {
                    "start": 211.2,
                    "end": 211.38,
                    "word": " had",
                    "probability": 0.80078125
                },
                {
                    "start": 211.38,
                    "end": 211.62,
                    "word": " grown",
                    "probability": 0.83349609375
                },
                {
                    "start": 211.62,
                    "end": 211.78,
                    "word": " to",
                    "probability": 0.8896484375
                },
                {
                    "start": 211.78,
                    "end": 212.06,
                    "word": " almost",
                    "probability": 0.69921875
                },
                {
                    "start": 212.06,
                    "end": 212.32,
                    "word": " six",
                    "probability": 0.4052734375
                },
                {
                    "start": 212.32,
                    "end": 212.64,
                    "word": " billion",
                    "probability": 0.80810546875
                },
                {
                    "start": 212.64,
                    "end": 213.12,
                    "word": ".",
                    "probability": 0.4423828125
                },
                {
                    "start": 213.26,
                    "end": 213.52,
                    "word": " And",
                    "probability": 0.8369140625
                },
                {
                    "start": 213.52,
                    "end": 213.98,
                    "word": " in",
                    "probability": 0.87939453125
                },
                {
                    "start": 213.98,
                    "end": 214.64,
                    "word": " 2018",
                    "probability": 0.74609375
                },
                {
                    "start": 214.64,
                    "end": 215.02,
                    "word": ",",
                    "probability": 0.85498046875
                }
            ]
        },
        {
            "id": 35,
            "seek": 21960,
            "start": 215.02,
            "end": 219.14,
            "text": " that number had grown to more than 12 billion. I don't have data for the last two years, but",
            "tokens": [
                50944,
                300,
                1230,
                632,
                7709,
                281,
                544,
                813,
                2272,
                5218,
                13,
                286,
                500,
                380,
                362,
                1412,
                337,
                264,
                1036,
                732,
                924,
                11,
                457,
                51172
            ],
            "temperature": 0.0,
            "avg_logprob": -0.17851217025149185,
            "compression_ratio": 1.6175438596491227,
            "no_speech_prob": 0.0006260871887207031,
            "words": [
                {
                    "start": 215.02,
                    "end": 215.2,
                    "word": " that",
                    "probability": 0.8798828125
                },
                {
                    "start": 215.2,
                    "end": 215.44,
                    "word": " number",
                    "probability": 0.87109375
                },
                {
                    "start": 215.44,
                    "end": 215.6,
                    "word": " had",
                    "probability": 0.796875
                },
                {
                    "start": 215.6,
                    "end": 215.76,
                    "word": " grown",
                    "probability": 0.8369140625
                },
                {
                    "start": 215.76,
                    "end": 215.92,
                    "word": " to",
                    "probability": 0.900390625
                },
                {
                    "start": 215.92,
                    "end": 216.04,
                    "word": " more",
                    "probability": 0.88818359375
                },
                {
                    "start": 216.04,
                    "end": 216.18,
                    "word": " than",
                    "probability": 0.87744140625
                },
                {
                    "start": 216.18,
                    "end": 216.42,
                    "word": " 12",
                    "probability": 0.84423828125
                },
                {
                    "start": 216.42,
                    "end": 216.7,
                    "word": " billion",
                    "probability": 0.80126953125
                },
                {
                    "start": 216.7,
                    "end": 217.18,
                    "word": ".",
                    "probability": 0.9921875
                },
                {
                    "start": 217.3,
                    "end": 217.56,
                    "word": " I",
                    "probability": 0.98486328125
                },
                {
                    "start": 217.56,
                    "end": 217.7,
                    "word": " don't",
                    "probability": 0.94384765625
                },
                {
                    "start": 217.7,
                    "end": 217.84,
                    "word": " have",
                    "probability": 0.87841796875
                },
                {
                    "start": 217.84,
                    "end": 218.0,
                    "word": " data",
                    "probability": 0.81787109375
                },
                {
                    "start": 218.0,
                    "end": 218.14,
                    "word": " for",
                    "probability": 0.88818359375
                },
                {
                    "start": 218.14,
                    "end": 218.26,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 218.26,
                    "end": 218.42,
                    "word": " last",
                    "probability": 0.8212890625
                },
                {
                    "start": 218.42,
                    "end": 218.62,
                    "word": " two",
                    "probability": 0.83056640625
                },
                {
                    "start": 218.62,
                    "end": 218.86,
                    "word": " years",
                    "probability": 0.87744140625
                },
                {
                    "start": 218.86,
                    "end": 218.96,
                    "word": ",",
                    "probability": 0.8076171875
                },
                {
                    "start": 218.96,
                    "end": 219.14,
                    "word": " but",
                    "probability": 0.89697265625
                }
            ]
        },
        {
            "id": 36,
            "seek": 21960,
            "start": 219.14,
            "end": 224.94,
            "text": " I see no reason to believe that this trend has attenuated. I think it's likely it's continued",
            "tokens": [
                51172,
                286,
                536,
                572,
                1778,
                281,
                1697,
                300,
                341,
                6028,
                575,
                951,
                268,
                27275,
                13,
                286,
                519,
                309,
                311,
                3700,
                309,
                311,
                7014,
                51448
            ],
            "temperature": 0.0,
            "avg_logprob": -0.17851217025149185,
            "compression_ratio": 1.6175438596491227,
            "no_speech_prob": 0.0006260871887207031,
            "words": [
                {
                    "start": 219.14,
                    "end": 219.64,
                    "word": " I",
                    "probability": 0.99560546875
                },
                {
                    "start": 219.64,
                    "end": 219.76,
                    "word": " see",
                    "probability": 0.8349609375
                },
                {
                    "start": 219.76,
                    "end": 219.92,
                    "word": " no",
                    "probability": 0.89794921875
                },
                {
                    "start": 219.92,
                    "end": 220.18,
                    "word": " reason",
                    "probability": 0.91162109375
                },
                {
                    "start": 220.18,
                    "end": 220.48,
                    "word": " to",
                    "probability": 0.90966796875
                },
                {
                    "start": 220.48,
                    "end": 220.82,
                    "word": " believe",
                    "probability": 0.72314453125
                },
                {
                    "start": 220.82,
                    "end": 221.2,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 221.2,
                    "end": 222.78,
                    "word": " this",
                    "probability": 0.70458984375
                },
                {
                    "start": 222.78,
                    "end": 223.02,
                    "word": " trend",
                    "probability": 0.81298828125
                },
                {
                    "start": 223.02,
                    "end": 223.16,
                    "word": " has",
                    "probability": 0.71240234375
                },
                {
                    "start": 223.16,
                    "end": 223.66,
                    "word": " attenuated",
                    "probability": 0.92578125
                },
                {
                    "start": 223.66,
                    "end": 223.88,
                    "word": ".",
                    "probability": 0.85546875
                },
                {
                    "start": 223.88,
                    "end": 224.02,
                    "word": " I",
                    "probability": 0.951171875
                },
                {
                    "start": 224.02,
                    "end": 224.12,
                    "word": " think",
                    "probability": 0.8154296875
                },
                {
                    "start": 224.12,
                    "end": 224.26,
                    "word": " it's",
                    "probability": 0.932373046875
                },
                {
                    "start": 224.26,
                    "end": 224.46,
                    "word": " likely",
                    "probability": 0.81103515625
                },
                {
                    "start": 224.46,
                    "end": 224.64,
                    "word": " it's",
                    "probability": 0.898193359375
                },
                {
                    "start": 224.64,
                    "end": 224.94,
                    "word": " continued",
                    "probability": 0.615234375
                }
            ]
        },
        {
            "id": 37,
            "seek": 21960,
            "start": 224.94,
            "end": 230.46,
            "text": " on a similar trajectory. This is a big problem. We see this in Sophos's customer base,",
            "tokens": [
                51448,
                322,
                257,
                2531,
                21512,
                13,
                639,
                307,
                257,
                955,
                1154,
                13,
                492,
                536,
                341,
                294,
                18921,
                329,
                311,
                5474,
                3096,
                11,
                51728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.17851217025149185,
            "compression_ratio": 1.6175438596491227,
            "no_speech_prob": 0.0006260871887207031,
            "words": [
                {
                    "start": 224.94,
                    "end": 225.16,
                    "word": " on",
                    "probability": 0.87646484375
                },
                {
                    "start": 225.24,
                    "end": 225.46,
                    "word": " a similar",
                    "probability": 0.93017578125
                },
                {
                    "start": 225.46,
                    "end": 225.92,
                    "word": " trajectory",
                    "probability": 0.80078125
                },
                {
                    "start": 225.92,
                    "end": 226.4,
                    "word": ".",
                    "probability": 0.99560546875
                },
                {
                    "start": 226.7,
                    "end": 226.98,
                    "word": " This",
                    "probability": 0.80224609375
                },
                {
                    "start": 226.98,
                    "end": 227.06,
                    "word": " is",
                    "probability": 0.90625
                },
                {
                    "start": 227.18,
                    "end": 227.3,
                    "word": " a big",
                    "probability": 0.89306640625
                },
                {
                    "start": 227.3,
                    "end": 227.64,
                    "word": " problem",
                    "probability": 0.76953125
                },
                {
                    "start": 227.64,
                    "end": 228.12,
                    "word": ".",
                    "probability": 0.97314453125
                },
                {
                    "start": 228.24,
                    "end": 228.34,
                    "word": " We",
                    "probability": 0.89453125
                },
                {
                    "start": 228.34,
                    "end": 228.48,
                    "word": " see",
                    "probability": 0.818359375
                },
                {
                    "start": 228.48,
                    "end": 228.66,
                    "word": " this",
                    "probability": 0.8759765625
                },
                {
                    "start": 228.66,
                    "end": 228.88,
                    "word": " in",
                    "probability": 0.87744140625
                },
                {
                    "start": 228.88,
                    "end": 229.54,
                    "word": " Sophos's",
                    "probability": 0.577392578125
                },
                {
                    "start": 229.54,
                    "end": 229.84,
                    "word": " customer",
                    "probability": 0.6787109375
                },
                {
                    "start": 229.84,
                    "end": 230.2,
                    "word": " base",
                    "probability": 0.81103515625
                },
                {
                    "start": 230.2,
                    "end": 230.46,
                    "word": ",",
                    "probability": 0.91650390625
                }
            ]
        },
        {
            "id": 38,
            "seek": 24700,
            "start": 230.46,
            "end": 235.52,
            "text": " which is which is a pretty large sample. We also hear about it from other folks in the",
            "tokens": [
                50368,
                597,
                307,
                597,
                307,
                257,
                1238,
                2416,
                6889,
                13,
                492,
                611,
                1568,
                466,
                309,
                490,
                661,
                4024,
                294,
                264,
                50608
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22982403433438645,
            "compression_ratio": 1.6133828996282529,
            "no_speech_prob": 0.012603759765625,
            "words": [
                {
                    "start": 230.46,
                    "end": 230.8,
                    "word": " which",
                    "probability": 0.36328125
                },
                {
                    "start": 230.8,
                    "end": 231.04,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 231.04,
                    "end": 231.32,
                    "word": " which",
                    "probability": 0.061553955078125
                },
                {
                    "start": 231.32,
                    "end": 231.44,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 231.54,
                    "end": 231.66,
                    "word": " a pretty",
                    "probability": 0.88037109375
                },
                {
                    "start": 231.66,
                    "end": 231.92,
                    "word": " large",
                    "probability": 0.92333984375
                },
                {
                    "start": 231.92,
                    "end": 232.3,
                    "word": " sample",
                    "probability": 0.77490234375
                },
                {
                    "start": 232.3,
                    "end": 232.78,
                    "word": ".",
                    "probability": 0.595703125
                },
                {
                    "start": 232.96,
                    "end": 233.06,
                    "word": " We",
                    "probability": 0.71337890625
                },
                {
                    "start": 233.06,
                    "end": 233.22,
                    "word": " also",
                    "probability": 0.794921875
                },
                {
                    "start": 233.22,
                    "end": 233.44,
                    "word": " hear",
                    "probability": 0.88671875
                },
                {
                    "start": 233.44,
                    "end": 233.66,
                    "word": " about",
                    "probability": 0.82177734375
                },
                {
                    "start": 233.66,
                    "end": 233.82,
                    "word": " it",
                    "probability": 0.89697265625
                },
                {
                    "start": 233.82,
                    "end": 234.1,
                    "word": " from",
                    "probability": 0.818359375
                },
                {
                    "start": 234.1,
                    "end": 235.0,
                    "word": " other",
                    "probability": 0.7744140625
                },
                {
                    "start": 235.0,
                    "end": 235.28,
                    "word": " folks",
                    "probability": 0.8330078125
                },
                {
                    "start": 235.28,
                    "end": 235.42,
                    "word": " in",
                    "probability": 0.7880859375
                },
                {
                    "start": 235.42,
                    "end": 235.52,
                    "word": " the",
                    "probability": 0.8056640625
                }
            ]
        },
        {
            "id": 39,
            "seek": 24700,
            "start": 235.52,
            "end": 239.94,
            "text": " cybersecurity space. And so we're very focused on it because it's affecting people. And it's",
            "tokens": [
                50608,
                38765,
                1901,
                13,
                400,
                370,
                321,
                434,
                588,
                5178,
                322,
                309,
                570,
                309,
                311,
                17476,
                561,
                13,
                400,
                309,
                311,
                50832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22982403433438645,
            "compression_ratio": 1.6133828996282529,
            "no_speech_prob": 0.012603759765625,
            "words": [
                {
                    "start": 235.52,
                    "end": 235.82,
                    "word": " cybersecurity",
                    "probability": 0.58544921875
                },
                {
                    "start": 235.82,
                    "end": 236.44,
                    "word": " space",
                    "probability": 0.8095703125
                },
                {
                    "start": 236.44,
                    "end": 236.92,
                    "word": ".",
                    "probability": 0.69921875
                },
                {
                    "start": 237.02,
                    "end": 237.24,
                    "word": " And",
                    "probability": 0.77685546875
                },
                {
                    "start": 237.24,
                    "end": 237.34,
                    "word": " so",
                    "probability": 0.90380859375
                },
                {
                    "start": 237.34,
                    "end": 237.46,
                    "word": " we're",
                    "probability": 0.912353515625
                },
                {
                    "start": 237.46,
                    "end": 237.64,
                    "word": " very",
                    "probability": 0.73779296875
                },
                {
                    "start": 237.64,
                    "end": 237.9,
                    "word": " focused",
                    "probability": 0.75439453125
                },
                {
                    "start": 237.9,
                    "end": 238.1,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 238.1,
                    "end": 238.24,
                    "word": " it",
                    "probability": 0.900390625
                },
                {
                    "start": 238.24,
                    "end": 238.4,
                    "word": " because",
                    "probability": 0.6669921875
                },
                {
                    "start": 238.4,
                    "end": 238.56,
                    "word": " it's",
                    "probability": 0.950927734375
                },
                {
                    "start": 238.56,
                    "end": 238.8,
                    "word": " affecting",
                    "probability": 0.7626953125
                },
                {
                    "start": 238.8,
                    "end": 239.22,
                    "word": " people",
                    "probability": 0.9169921875
                },
                {
                    "start": 239.22,
                    "end": 239.7,
                    "word": ".",
                    "probability": 0.7568359375
                },
                {
                    "start": 239.74,
                    "end": 239.82,
                    "word": " And",
                    "probability": 0.476806640625
                },
                {
                    "start": 239.82,
                    "end": 239.94,
                    "word": " it's",
                    "probability": 0.951171875
                }
            ]
        },
        {
            "id": 40,
            "seek": 24700,
            "start": 239.94,
            "end": 244.22,
            "text": " not just affecting large organizations. You can see on this on the axis on the right,",
            "tokens": [
                50832,
                406,
                445,
                17476,
                2416,
                6150,
                13,
                509,
                393,
                536,
                322,
                341,
                322,
                264,
                10298,
                322,
                264,
                558,
                11,
                51016
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22982403433438645,
            "compression_ratio": 1.6133828996282529,
            "no_speech_prob": 0.012603759765625,
            "words": [
                {
                    "start": 239.94,
                    "end": 240.06,
                    "word": " not",
                    "probability": 0.8994140625
                },
                {
                    "start": 240.06,
                    "end": 240.2,
                    "word": " just",
                    "probability": 0.88916015625
                },
                {
                    "start": 240.2,
                    "end": 240.52,
                    "word": " affecting",
                    "probability": 0.75830078125
                },
                {
                    "start": 240.52,
                    "end": 241.02,
                    "word": " large",
                    "probability": 0.92626953125
                },
                {
                    "start": 241.02,
                    "end": 241.64,
                    "word": " organizations",
                    "probability": 0.7060546875
                },
                {
                    "start": 241.64,
                    "end": 241.82,
                    "word": ".",
                    "probability": 0.7607421875
                },
                {
                    "start": 241.82,
                    "end": 241.86,
                    "word": " You",
                    "probability": 0.88916015625
                },
                {
                    "start": 241.86,
                    "end": 241.96,
                    "word": " can",
                    "probability": 0.89892578125
                },
                {
                    "start": 241.96,
                    "end": 242.1,
                    "word": " see",
                    "probability": 0.8115234375
                },
                {
                    "start": 242.1,
                    "end": 242.24,
                    "word": " on",
                    "probability": 0.88037109375
                },
                {
                    "start": 242.24,
                    "end": 242.4,
                    "word": " this",
                    "probability": 0.771484375
                },
                {
                    "start": 242.4,
                    "end": 242.62,
                    "word": " on",
                    "probability": 0.7177734375
                },
                {
                    "start": 242.62,
                    "end": 242.76,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 242.76,
                    "end": 243.0,
                    "word": " axis",
                    "probability": 0.869140625
                },
                {
                    "start": 243.0,
                    "end": 243.22,
                    "word": " on",
                    "probability": 0.87255859375
                },
                {
                    "start": 243.22,
                    "end": 243.34,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 243.34,
                    "end": 243.58,
                    "word": " right",
                    "probability": 0.88525390625
                },
                {
                    "start": 243.58,
                    "end": 244.22,
                    "word": ",",
                    "probability": 0.2174072265625
                }
            ]
        },
        {
            "id": 41,
            "seek": 24700,
            "start": 244.22,
            "end": 252.58,
            "text": " that something like 80,000 organizations had been hit by July 2018 by these attacks.",
            "tokens": [
                51044,
                300,
                746,
                411,
                4688,
                11,
                1360,
                6150,
                632,
                668,
                2045,
                538,
                7370,
                6096,
                538,
                613,
                8122,
                13,
                51444
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22982403433438645,
            "compression_ratio": 1.6133828996282529,
            "no_speech_prob": 0.012603759765625,
            "words": [
                {
                    "start": 244.22,
                    "end": 244.48,
                    "word": " that",
                    "probability": 0.833984375
                },
                {
                    "start": 244.48,
                    "end": 244.88,
                    "word": " something",
                    "probability": 0.51806640625
                },
                {
                    "start": 244.88,
                    "end": 245.52,
                    "word": " like",
                    "probability": 0.87548828125
                },
                {
                    "start": 245.52,
                    "end": 247.06,
                    "word": " 80",
                    "probability": 0.60888671875
                },
                {
                    "start": 247.06,
                    "end": 247.34,
                    "word": ",000",
                    "probability": 0.976318359375
                },
                {
                    "start": 247.34,
                    "end": 247.96,
                    "word": " organizations",
                    "probability": 0.69091796875
                },
                {
                    "start": 247.96,
                    "end": 248.82,
                    "word": " had",
                    "probability": 0.75244140625
                },
                {
                    "start": 248.82,
                    "end": 248.98,
                    "word": " been",
                    "probability": 0.81787109375
                },
                {
                    "start": 248.98,
                    "end": 249.22,
                    "word": " hit",
                    "probability": 0.904296875
                },
                {
                    "start": 249.22,
                    "end": 249.58,
                    "word": " by",
                    "probability": 0.90185546875
                },
                {
                    "start": 249.58,
                    "end": 250.12,
                    "word": " July",
                    "probability": 0.1385498046875
                },
                {
                    "start": 250.12,
                    "end": 250.78,
                    "word": " 2018",
                    "probability": 0.64013671875
                },
                {
                    "start": 250.78,
                    "end": 251.38,
                    "word": " by",
                    "probability": 0.794921875
                },
                {
                    "start": 251.38,
                    "end": 251.64,
                    "word": " these",
                    "probability": 0.7041015625
                },
                {
                    "start": 251.64,
                    "end": 252.1,
                    "word": " attacks",
                    "probability": 0.7197265625
                },
                {
                    "start": 252.1,
                    "end": 252.58,
                    "word": ".",
                    "probability": 0.9892578125
                }
            ]
        },
        {
            "id": 42,
            "seek": 24700,
            "start": 253.9,
            "end": 257.86,
            "text": " So there are lots of small and mid-sized organizations getting hit. And oftentimes,",
            "tokens": [
                51548,
                407,
                456,
                366,
                3195,
                295,
                1359,
                293,
                2062,
                12,
                20614,
                6150,
                1242,
                2045,
                13,
                400,
                18349,
                11,
                51724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22982403433438645,
            "compression_ratio": 1.6133828996282529,
            "no_speech_prob": 0.012603759765625,
            "words": [
                {
                    "start": 253.9,
                    "end": 254.38,
                    "word": " So",
                    "probability": 0.8876953125
                },
                {
                    "start": 254.38,
                    "end": 254.56,
                    "word": " there",
                    "probability": 0.830078125
                },
                {
                    "start": 254.56,
                    "end": 254.64,
                    "word": " are",
                    "probability": 0.888671875
                },
                {
                    "start": 254.64,
                    "end": 254.88,
                    "word": " lots",
                    "probability": 0.8095703125
                },
                {
                    "start": 254.88,
                    "end": 255.04,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 255.04,
                    "end": 255.3,
                    "word": " small",
                    "probability": 0.84228515625
                },
                {
                    "start": 255.3,
                    "end": 255.44,
                    "word": " and",
                    "probability": 0.87451171875
                },
                {
                    "start": 255.44,
                    "end": 255.6,
                    "word": " mid",
                    "probability": 0.8876953125
                },
                {
                    "start": 255.6,
                    "end": 255.78,
                    "word": "-sized",
                    "probability": 0.4630126953125
                },
                {
                    "start": 255.78,
                    "end": 256.32,
                    "word": " organizations",
                    "probability": 0.7099609375
                },
                {
                    "start": 256.32,
                    "end": 256.66,
                    "word": " getting",
                    "probability": 0.9013671875
                },
                {
                    "start": 256.66,
                    "end": 256.9,
                    "word": " hit",
                    "probability": 0.87060546875
                },
                {
                    "start": 256.9,
                    "end": 256.98,
                    "word": ".",
                    "probability": 0.456787109375
                },
                {
                    "start": 256.98,
                    "end": 257.08,
                    "word": " And",
                    "probability": 0.87060546875
                },
                {
                    "start": 257.08,
                    "end": 257.38,
                    "word": " oftentimes",
                    "probability": 0.80859375
                },
                {
                    "start": 257.38,
                    "end": 257.86,
                    "word": ",",
                    "probability": 0.267578125
                }
            ]
        },
        {
            "id": 43,
            "seek": 27470,
            "start": 257.86,
            "end": 263.1,
            "text": " financial damage can be in the hundreds of thousands of dollars and really impact people's",
            "tokens": [
                50364,
                4669,
                4344,
                393,
                312,
                294,
                264,
                6779,
                295,
                5383,
                295,
                3808,
                293,
                534,
                2712,
                561,
                311,
                50620
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22819010075181723,
            "compression_ratio": 1.6415094339622642,
            "no_speech_prob": 0.0010633468627929688,
            "words": [
                {
                    "start": 257.86,
                    "end": 258.34,
                    "word": " financial",
                    "probability": 0.158447265625
                },
                {
                    "start": 258.34,
                    "end": 258.88,
                    "word": " damage",
                    "probability": 0.7587890625
                },
                {
                    "start": 258.88,
                    "end": 259.26,
                    "word": " can",
                    "probability": 0.83203125
                },
                {
                    "start": 259.26,
                    "end": 259.46,
                    "word": " be",
                    "probability": 0.8994140625
                },
                {
                    "start": 259.46,
                    "end": 259.64,
                    "word": " in",
                    "probability": 0.81201171875
                },
                {
                    "start": 259.64,
                    "end": 259.82,
                    "word": " the",
                    "probability": 0.63720703125
                },
                {
                    "start": 259.82,
                    "end": 260.08,
                    "word": " hundreds",
                    "probability": 0.77783203125
                },
                {
                    "start": 260.08,
                    "end": 260.74,
                    "word": " of",
                    "probability": 0.74658203125
                },
                {
                    "start": 260.74,
                    "end": 260.94,
                    "word": " thousands",
                    "probability": 0.68408203125
                },
                {
                    "start": 260.94,
                    "end": 261.12,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 261.12,
                    "end": 261.34,
                    "word": " dollars",
                    "probability": 0.92822265625
                },
                {
                    "start": 261.34,
                    "end": 261.78,
                    "word": " and",
                    "probability": 0.783203125
                },
                {
                    "start": 261.78,
                    "end": 262.26,
                    "word": " really",
                    "probability": 0.7099609375
                },
                {
                    "start": 262.26,
                    "end": 262.7,
                    "word": " impact",
                    "probability": 0.75537109375
                },
                {
                    "start": 262.7,
                    "end": 263.1,
                    "word": " people's",
                    "probability": 0.95068359375
                }
            ]
        },
        {
            "id": 44,
            "seek": 27470,
            "start": 263.1,
            "end": 269.24,
            "text": " lives when these attacks happened. So again, just to reiterate, we're focused on primarily",
            "tokens": [
                50620,
                2909,
                562,
                613,
                8122,
                2011,
                13,
                407,
                797,
                11,
                445,
                281,
                33528,
                11,
                321,
                434,
                5178,
                322,
                10029,
                50932
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22819010075181723,
            "compression_ratio": 1.6415094339622642,
            "no_speech_prob": 0.0010633468627929688,
            "words": [
                {
                    "start": 263.1,
                    "end": 263.3,
                    "word": " lives",
                    "probability": 0.7529296875
                },
                {
                    "start": 263.3,
                    "end": 263.54,
                    "word": " when",
                    "probability": 0.8681640625
                },
                {
                    "start": 263.54,
                    "end": 263.9,
                    "word": " these",
                    "probability": 0.59033203125
                },
                {
                    "start": 263.9,
                    "end": 264.18,
                    "word": " attacks",
                    "probability": 0.70458984375
                },
                {
                    "start": 264.18,
                    "end": 264.54,
                    "word": " happened",
                    "probability": 0.41845703125
                },
                {
                    "start": 264.54,
                    "end": 265.06,
                    "word": ".",
                    "probability": 0.8837890625
                },
                {
                    "start": 265.06,
                    "end": 265.06,
                    "word": " So",
                    "probability": 0.73779296875
                },
                {
                    "start": 265.06,
                    "end": 267.08,
                    "word": " again",
                    "probability": 0.615234375
                },
                {
                    "start": 267.08,
                    "end": 267.18,
                    "word": ",",
                    "probability": 0.61474609375
                },
                {
                    "start": 267.18,
                    "end": 267.3,
                    "word": " just",
                    "probability": 0.8798828125
                },
                {
                    "start": 267.3,
                    "end": 267.44,
                    "word": " to",
                    "probability": 0.90771484375
                },
                {
                    "start": 267.44,
                    "end": 267.68,
                    "word": " reiterate",
                    "probability": 0.880859375
                },
                {
                    "start": 267.68,
                    "end": 267.8,
                    "word": ",",
                    "probability": 0.76953125
                },
                {
                    "start": 267.8,
                    "end": 267.9,
                    "word": " we're",
                    "probability": 0.90283203125
                },
                {
                    "start": 267.9,
                    "end": 268.16,
                    "word": " focused",
                    "probability": 0.7685546875
                },
                {
                    "start": 268.16,
                    "end": 268.5,
                    "word": " on",
                    "probability": 0.884765625
                },
                {
                    "start": 268.5,
                    "end": 269.24,
                    "word": " primarily",
                    "probability": 0.81201171875
                }
            ]
        },
        {
            "id": 45,
            "seek": 27470,
            "start": 269.24,
            "end": 273.22,
            "text": " these business email compromise use cases, but also more generally targeted phishing in which",
            "tokens": [
                50932,
                613,
                1606,
                3796,
                18577,
                764,
                3331,
                11,
                457,
                611,
                544,
                5101,
                15045,
                903,
                3807,
                294,
                597,
                51124
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22819010075181723,
            "compression_ratio": 1.6415094339622642,
            "no_speech_prob": 0.0010633468627929688,
            "words": [
                {
                    "start": 269.24,
                    "end": 269.52,
                    "word": " these",
                    "probability": 0.6904296875
                },
                {
                    "start": 269.52,
                    "end": 269.8,
                    "word": " business",
                    "probability": 0.86669921875
                },
                {
                    "start": 269.8,
                    "end": 270.0,
                    "word": " email",
                    "probability": 0.85546875
                },
                {
                    "start": 270.0,
                    "end": 270.34,
                    "word": " compromise",
                    "probability": 0.421875
                },
                {
                    "start": 270.34,
                    "end": 270.64,
                    "word": " use",
                    "probability": 0.80908203125
                },
                {
                    "start": 270.64,
                    "end": 270.96,
                    "word": " cases",
                    "probability": 0.81640625
                },
                {
                    "start": 270.96,
                    "end": 271.04,
                    "word": ",",
                    "probability": 0.5478515625
                },
                {
                    "start": 271.04,
                    "end": 271.18,
                    "word": " but",
                    "probability": 0.89990234375
                },
                {
                    "start": 271.18,
                    "end": 271.4,
                    "word": " also",
                    "probability": 0.82275390625
                },
                {
                    "start": 271.4,
                    "end": 271.6,
                    "word": " more",
                    "probability": 0.83154296875
                },
                {
                    "start": 271.6,
                    "end": 271.94,
                    "word": " generally",
                    "probability": 0.82958984375
                },
                {
                    "start": 271.94,
                    "end": 272.4,
                    "word": " targeted",
                    "probability": 0.71142578125
                },
                {
                    "start": 272.4,
                    "end": 272.74,
                    "word": " phishing",
                    "probability": 0.89208984375
                },
                {
                    "start": 272.74,
                    "end": 272.92,
                    "word": " in",
                    "probability": 0.541015625
                },
                {
                    "start": 272.92,
                    "end": 273.22,
                    "word": " which",
                    "probability": 0.916015625
                }
            ]
        },
        {
            "id": 46,
            "seek": 27470,
            "start": 274.6,
            "end": 278.92,
            "text": " a lot of manual labor goes into the phishing process on the criminal actors.",
            "tokens": [
                51192,
                257,
                688,
                295,
                9688,
                5938,
                1709,
                666,
                264,
                903,
                3807,
                1399,
                322,
                264,
                8628,
                10037,
                13,
                51392
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22819010075181723,
            "compression_ratio": 1.6415094339622642,
            "no_speech_prob": 0.0010633468627929688,
            "words": [
                {
                    "start": 274.6,
                    "end": 274.76,
                    "word": " a lot",
                    "probability": 0.8896484375
                },
                {
                    "start": 274.76,
                    "end": 274.9,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 274.9,
                    "end": 275.2,
                    "word": " manual",
                    "probability": 0.92529296875
                },
                {
                    "start": 275.2,
                    "end": 275.44,
                    "word": " labor",
                    "probability": 0.72802734375
                },
                {
                    "start": 275.44,
                    "end": 275.72,
                    "word": " goes",
                    "probability": 0.859375
                },
                {
                    "start": 275.72,
                    "end": 276.12,
                    "word": " into",
                    "probability": 0.72314453125
                },
                {
                    "start": 276.12,
                    "end": 276.96,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 276.96,
                    "end": 277.16,
                    "word": " phishing",
                    "probability": 0.91259765625
                },
                {
                    "start": 277.16,
                    "end": 277.56,
                    "word": " process",
                    "probability": 0.85888671875
                },
                {
                    "start": 277.56,
                    "end": 277.74,
                    "word": " on",
                    "probability": 0.90576171875
                },
                {
                    "start": 277.74,
                    "end": 277.84,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 277.84,
                    "end": 278.06,
                    "word": " criminal",
                    "probability": 0.76025390625
                },
                {
                    "start": 278.06,
                    "end": 278.4,
                    "word": " actors",
                    "probability": 0.72119140625
                },
                {
                    "start": 278.4,
                    "end": 278.92,
                    "word": ".",
                    "probability": 0.9560546875
                }
            ]
        },
        {
            "id": 47,
            "seek": 27470,
            "start": 281.52,
            "end": 285.56,
            "text": " And I think it almost goes without saying, but we're focused on step two and three",
            "tokens": [
                51560,
                400,
                286,
                519,
                309,
                1920,
                1709,
                1553,
                1566,
                11,
                457,
                321,
                434,
                5178,
                322,
                1823,
                732,
                293,
                1045,
                51756
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22819010075181723,
            "compression_ratio": 1.6415094339622642,
            "no_speech_prob": 0.0010633468627929688,
            "words": [
                {
                    "start": 281.52,
                    "end": 282.04,
                    "word": " And",
                    "probability": 0.67529296875
                },
                {
                    "start": 282.04,
                    "end": 282.14,
                    "word": " I",
                    "probability": 0.98388671875
                },
                {
                    "start": 282.14,
                    "end": 282.26,
                    "word": " think",
                    "probability": 0.81396484375
                },
                {
                    "start": 282.26,
                    "end": 282.54,
                    "word": " it",
                    "probability": 0.90087890625
                },
                {
                    "start": 282.54,
                    "end": 282.76,
                    "word": " almost",
                    "probability": 0.6572265625
                },
                {
                    "start": 282.76,
                    "end": 282.98,
                    "word": " goes",
                    "probability": 0.857421875
                },
                {
                    "start": 282.98,
                    "end": 283.18,
                    "word": " without",
                    "probability": 0.8193359375
                },
                {
                    "start": 283.18,
                    "end": 283.46,
                    "word": " saying",
                    "probability": 0.82275390625
                },
                {
                    "start": 283.46,
                    "end": 283.54,
                    "word": ",",
                    "probability": 0.62255859375
                },
                {
                    "start": 283.54,
                    "end": 283.6,
                    "word": " but",
                    "probability": 0.83740234375
                },
                {
                    "start": 283.6,
                    "end": 283.74,
                    "word": " we're",
                    "probability": 0.912353515625
                },
                {
                    "start": 283.74,
                    "end": 284.3,
                    "word": " focused",
                    "probability": 0.68017578125
                },
                {
                    "start": 284.3,
                    "end": 284.54,
                    "word": " on",
                    "probability": 0.91162109375
                },
                {
                    "start": 284.54,
                    "end": 284.8,
                    "word": " step",
                    "probability": 0.69140625
                },
                {
                    "start": 284.8,
                    "end": 285.1,
                    "word": " two",
                    "probability": 0.71337890625
                },
                {
                    "start": 285.1,
                    "end": 285.26,
                    "word": " and",
                    "probability": 0.88818359375
                },
                {
                    "start": 285.26,
                    "end": 285.56,
                    "word": " three",
                    "probability": 0.84326171875
                }
            ]
        },
        {
            "id": 48,
            "seek": 30470,
            "start": 285.56,
            "end": 291.3,
            "text": " of these criminal actors workflow, the steps that are mediated over email. So we see lots of",
            "tokens": [
                50392,
                295,
                613,
                8628,
                10037,
                20993,
                11,
                264,
                4439,
                300,
                366,
                17269,
                770,
                670,
                3796,
                13,
                407,
                321,
                536,
                3195,
                295,
                50644
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23874999850988388,
            "compression_ratio": 1.703125,
            "no_speech_prob": 0.0012054443359375,
            "words": [
                {
                    "start": 285.56,
                    "end": 286.02,
                    "word": " of",
                    "probability": 0.73486328125
                },
                {
                    "start": 286.02,
                    "end": 286.82,
                    "word": " these",
                    "probability": 0.654296875
                },
                {
                    "start": 286.82,
                    "end": 287.08,
                    "word": " criminal",
                    "probability": 0.78369140625
                },
                {
                    "start": 287.08,
                    "end": 287.38,
                    "word": " actors",
                    "probability": 0.7333984375
                },
                {
                    "start": 287.38,
                    "end": 287.84,
                    "word": " workflow",
                    "probability": 0.53173828125
                },
                {
                    "start": 287.84,
                    "end": 288.4,
                    "word": ",",
                    "probability": 0.7236328125
                },
                {
                    "start": 288.4,
                    "end": 288.66,
                    "word": " the",
                    "probability": 0.80517578125
                },
                {
                    "start": 288.66,
                    "end": 288.9,
                    "word": " steps",
                    "probability": 0.7119140625
                },
                {
                    "start": 288.9,
                    "end": 289.04,
                    "word": " that",
                    "probability": 0.87841796875
                },
                {
                    "start": 289.04,
                    "end": 289.14,
                    "word": " are",
                    "probability": 0.8857421875
                },
                {
                    "start": 289.14,
                    "end": 289.48,
                    "word": " mediated",
                    "probability": 0.8662109375
                },
                {
                    "start": 289.48,
                    "end": 289.68,
                    "word": " over",
                    "probability": 0.8798828125
                },
                {
                    "start": 289.68,
                    "end": 289.94,
                    "word": " email",
                    "probability": 0.82177734375
                },
                {
                    "start": 289.94,
                    "end": 290.16,
                    "word": ".",
                    "probability": 0.70361328125
                },
                {
                    "start": 290.16,
                    "end": 290.16,
                    "word": " So",
                    "probability": 0.845703125
                },
                {
                    "start": 290.16,
                    "end": 290.7,
                    "word": " we",
                    "probability": 0.72216796875
                },
                {
                    "start": 290.7,
                    "end": 290.86,
                    "word": " see",
                    "probability": 0.8056640625
                },
                {
                    "start": 290.86,
                    "end": 291.1,
                    "word": " lots",
                    "probability": 0.810546875
                },
                {
                    "start": 291.1,
                    "end": 291.3,
                    "word": " of",
                    "probability": 0.90673828125
                }
            ]
        },
        {
            "id": 49,
            "seek": 30470,
            "start": 291.3,
            "end": 298.84,
            "text": " malicious emails exchanged in the grooming step. And then we also see, obviously, a malicious email",
            "tokens": [
                50644,
                33496,
                12524,
                38378,
                294,
                264,
                49700,
                1823,
                13,
                400,
                550,
                321,
                611,
                536,
                11,
                2745,
                11,
                257,
                33496,
                3796,
                51028
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23874999850988388,
            "compression_ratio": 1.703125,
            "no_speech_prob": 0.0012054443359375,
            "words": [
                {
                    "start": 291.3,
                    "end": 291.56,
                    "word": " malicious",
                    "probability": 0.88525390625
                },
                {
                    "start": 291.56,
                    "end": 291.94,
                    "word": " emails",
                    "probability": 0.732421875
                },
                {
                    "start": 291.94,
                    "end": 294.36,
                    "word": " exchanged",
                    "probability": 0.74951171875
                },
                {
                    "start": 294.36,
                    "end": 294.78,
                    "word": " in",
                    "probability": 0.6953125
                },
                {
                    "start": 294.78,
                    "end": 294.98,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 294.98,
                    "end": 295.2,
                    "word": " grooming",
                    "probability": 0.85546875
                },
                {
                    "start": 295.2,
                    "end": 295.6,
                    "word": " step",
                    "probability": 0.8251953125
                },
                {
                    "start": 295.6,
                    "end": 295.72,
                    "word": ".",
                    "probability": 0.51220703125
                },
                {
                    "start": 295.72,
                    "end": 295.86,
                    "word": " And",
                    "probability": 0.892578125
                },
                {
                    "start": 295.86,
                    "end": 295.94,
                    "word": " then",
                    "probability": 0.73193359375
                },
                {
                    "start": 295.94,
                    "end": 296.04,
                    "word": " we",
                    "probability": 0.8955078125
                },
                {
                    "start": 296.04,
                    "end": 296.26,
                    "word": " also",
                    "probability": 0.81494140625
                },
                {
                    "start": 296.26,
                    "end": 296.7,
                    "word": " see",
                    "probability": 0.8193359375
                },
                {
                    "start": 296.7,
                    "end": 296.92,
                    "word": ",",
                    "probability": 0.333251953125
                },
                {
                    "start": 296.92,
                    "end": 298.08,
                    "word": " obviously",
                    "probability": 0.7578125
                },
                {
                    "start": 298.08,
                    "end": 298.26,
                    "word": ",",
                    "probability": 0.58349609375
                },
                {
                    "start": 298.3,
                    "end": 298.56,
                    "word": " a malicious",
                    "probability": 0.86279296875
                },
                {
                    "start": 298.56,
                    "end": 298.84,
                    "word": " email",
                    "probability": 0.892578125
                }
            ]
        },
        {
            "id": 50,
            "seek": 30470,
            "start": 298.84,
            "end": 302.44,
            "text": " transmitted in the exchange of information step where the attacker makes the ask of the",
            "tokens": [
                51028,
                25355,
                294,
                264,
                7742,
                295,
                1589,
                1823,
                689,
                264,
                35871,
                1669,
                264,
                1029,
                295,
                264,
                51196
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23874999850988388,
            "compression_ratio": 1.703125,
            "no_speech_prob": 0.0012054443359375,
            "words": [
                {
                    "start": 298.84,
                    "end": 299.3,
                    "word": " transmitted",
                    "probability": 0.85595703125
                },
                {
                    "start": 299.3,
                    "end": 299.5,
                    "word": " in",
                    "probability": 0.8828125
                },
                {
                    "start": 299.5,
                    "end": 299.6,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 299.6,
                    "end": 299.86,
                    "word": " exchange",
                    "probability": 0.7998046875
                },
                {
                    "start": 299.86,
                    "end": 299.96,
                    "word": " of",
                    "probability": 0.89306640625
                },
                {
                    "start": 299.96,
                    "end": 300.32,
                    "word": " information",
                    "probability": 0.734375
                },
                {
                    "start": 300.32,
                    "end": 300.7,
                    "word": " step",
                    "probability": 0.8173828125
                },
                {
                    "start": 300.7,
                    "end": 300.86,
                    "word": " where",
                    "probability": 0.58837890625
                },
                {
                    "start": 300.86,
                    "end": 301.02,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 301.02,
                    "end": 301.24,
                    "word": " attacker",
                    "probability": 0.72021484375
                },
                {
                    "start": 301.24,
                    "end": 301.54,
                    "word": " makes",
                    "probability": 0.71630859375
                },
                {
                    "start": 301.54,
                    "end": 301.68,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 301.68,
                    "end": 301.94,
                    "word": " ask",
                    "probability": 0.80322265625
                },
                {
                    "start": 301.94,
                    "end": 302.12,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 302.12,
                    "end": 302.44,
                    "word": " the",
                    "probability": 0.80859375
                }
            ]
        },
        {
            "id": 51,
            "seek": 30470,
            "start": 303.04,
            "end": 310.92,
            "text": " target, usually employee. And this is just because we're focused on email as our signal.",
            "tokens": [
                51244,
                3779,
                11,
                2673,
                10738,
                13,
                400,
                341,
                307,
                445,
                570,
                321,
                434,
                5178,
                322,
                3796,
                382,
                527,
                6358,
                13,
                51612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23874999850988388,
            "compression_ratio": 1.703125,
            "no_speech_prob": 0.0012054443359375,
            "words": [
                {
                    "start": 303.04,
                    "end": 303.5,
                    "word": " target",
                    "probability": 0.78759765625
                },
                {
                    "start": 303.5,
                    "end": 303.7,
                    "word": ",",
                    "probability": 0.87255859375
                },
                {
                    "start": 303.7,
                    "end": 303.9,
                    "word": " usually",
                    "probability": 0.82763671875
                },
                {
                    "start": 303.9,
                    "end": 304.36,
                    "word": " employee",
                    "probability": 0.63232421875
                },
                {
                    "start": 304.36,
                    "end": 304.82,
                    "word": ".",
                    "probability": 0.94580078125
                },
                {
                    "start": 307.52,
                    "end": 307.98,
                    "word": " And",
                    "probability": 0.87548828125
                },
                {
                    "start": 307.98,
                    "end": 308.16,
                    "word": " this",
                    "probability": 0.88427734375
                },
                {
                    "start": 308.16,
                    "end": 308.26,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 308.26,
                    "end": 308.4,
                    "word": " just",
                    "probability": 0.88232421875
                },
                {
                    "start": 308.4,
                    "end": 308.64,
                    "word": " because",
                    "probability": 0.8447265625
                },
                {
                    "start": 308.64,
                    "end": 308.8,
                    "word": " we're",
                    "probability": 0.923828125
                },
                {
                    "start": 308.8,
                    "end": 309.06,
                    "word": " focused",
                    "probability": 0.771484375
                },
                {
                    "start": 309.06,
                    "end": 309.22,
                    "word": " on",
                    "probability": 0.90625
                },
                {
                    "start": 309.22,
                    "end": 309.46,
                    "word": " email",
                    "probability": 0.90283203125
                },
                {
                    "start": 309.46,
                    "end": 309.68,
                    "word": " as",
                    "probability": 0.82177734375
                },
                {
                    "start": 309.68,
                    "end": 310.14,
                    "word": " our",
                    "probability": 0.806640625
                },
                {
                    "start": 310.14,
                    "end": 310.46,
                    "word": " signal",
                    "probability": 0.70751953125
                },
                {
                    "start": 310.46,
                    "end": 310.92,
                    "word": ".",
                    "probability": 0.76806640625
                }
            ]
        },
        {
            "id": 52,
            "seek": 30470,
            "start": 311.86,
            "end": 314.5,
            "text": " And that's the scope of the work that I'll be talking about today.",
            "tokens": [
                51684,
                400,
                300,
                311,
                264,
                11923,
                295,
                264,
                589,
                300,
                286,
                603,
                312,
                1417,
                466,
                965,
                13,
                51804
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23874999850988388,
            "compression_ratio": 1.703125,
            "no_speech_prob": 0.0012054443359375,
            "words": [
                {
                    "start": 311.86,
                    "end": 312.32,
                    "word": " And",
                    "probability": 0.8671875
                },
                {
                    "start": 312.32,
                    "end": 312.64,
                    "word": " that's",
                    "probability": 0.93896484375
                },
                {
                    "start": 312.64,
                    "end": 312.8,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 312.8,
                    "end": 312.96,
                    "word": " scope",
                    "probability": 0.8544921875
                },
                {
                    "start": 312.96,
                    "end": 313.1,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 313.1,
                    "end": 313.2,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 313.2,
                    "end": 313.34,
                    "word": " work",
                    "probability": 0.81005859375
                },
                {
                    "start": 313.34,
                    "end": 313.46,
                    "word": " that",
                    "probability": 0.88232421875
                },
                {
                    "start": 313.46,
                    "end": 313.52,
                    "word": " I'll",
                    "probability": 0.963623046875
                },
                {
                    "start": 313.52,
                    "end": 313.62,
                    "word": " be",
                    "probability": 0.90283203125
                },
                {
                    "start": 313.62,
                    "end": 313.84,
                    "word": " talking",
                    "probability": 0.76220703125
                },
                {
                    "start": 313.84,
                    "end": 314.12,
                    "word": " about",
                    "probability": 0.82373046875
                },
                {
                    "start": 314.12,
                    "end": 314.36,
                    "word": " today",
                    "probability": 0.79541015625
                },
                {
                    "start": 314.36,
                    "end": 314.5,
                    "word": ".",
                    "probability": 0.9990234375
                }
            ]
        },
        {
            "id": 53,
            "seek": 33178,
            "start": 316.18,
            "end": 321.76,
            "text": " Just to flesh this out a little bit, here's an example, phishing email sent in the later",
            "tokens": [
                50408,
                1449,
                281,
                12497,
                341,
                484,
                257,
                707,
                857,
                11,
                510,
                311,
                364,
                1365,
                11,
                903,
                3807,
                3796,
                2279,
                294,
                264,
                1780,
                50668
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2376838298404918,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0005435943603515625,
            "words": [
                {
                    "start": 316.18,
                    "end": 316.76,
                    "word": " Just",
                    "probability": 0.71142578125
                },
                {
                    "start": 316.76,
                    "end": 316.88,
                    "word": " to",
                    "probability": 0.9013671875
                },
                {
                    "start": 316.88,
                    "end": 317.04,
                    "word": " flesh",
                    "probability": 0.7509765625
                },
                {
                    "start": 317.04,
                    "end": 317.18,
                    "word": " this",
                    "probability": 0.8779296875
                },
                {
                    "start": 317.18,
                    "end": 317.34,
                    "word": " out",
                    "probability": 0.82470703125
                },
                {
                    "start": 317.42,
                    "end": 317.5,
                    "word": " a little",
                    "probability": 0.80810546875
                },
                {
                    "start": 317.5,
                    "end": 317.64,
                    "word": " bit",
                    "probability": 0.89306640625
                },
                {
                    "start": 317.64,
                    "end": 317.68,
                    "word": ",",
                    "probability": 0.7607421875
                },
                {
                    "start": 317.68,
                    "end": 317.84,
                    "word": " here's",
                    "probability": 0.881591796875
                },
                {
                    "start": 317.84,
                    "end": 317.94,
                    "word": " an",
                    "probability": 0.9052734375
                },
                {
                    "start": 317.94,
                    "end": 318.4,
                    "word": " example",
                    "probability": 0.93212890625
                },
                {
                    "start": 318.4,
                    "end": 319.24,
                    "word": ",",
                    "probability": 0.2166748046875
                },
                {
                    "start": 319.24,
                    "end": 319.52,
                    "word": " phishing",
                    "probability": 0.732177734375
                },
                {
                    "start": 319.52,
                    "end": 319.88,
                    "word": " email",
                    "probability": 0.80810546875
                },
                {
                    "start": 319.88,
                    "end": 320.42,
                    "word": " sent",
                    "probability": 0.379150390625
                },
                {
                    "start": 320.42,
                    "end": 320.8,
                    "word": " in",
                    "probability": 0.88232421875
                },
                {
                    "start": 320.8,
                    "end": 321.2,
                    "word": " the",
                    "probability": 0.80078125
                },
                {
                    "start": 321.2,
                    "end": 321.76,
                    "word": " later",
                    "probability": 0.8837890625
                }
            ]
        },
        {
            "id": 54,
            "seek": 33178,
            "start": 324.06,
            "end": 330.9,
            "text": " epics of the grooming stage of an attacker's workflow. Here, the attacker has established",
            "tokens": [
                50796,
                2388,
                1167,
                295,
                264,
                49700,
                3233,
                295,
                364,
                35871,
                311,
                20993,
                13,
                1692,
                11,
                264,
                35871,
                575,
                7545,
                51132
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2376838298404918,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0005435943603515625,
            "words": [
                {
                    "start": 324.06,
                    "end": 324.64,
                    "word": " epics",
                    "probability": 0.7734375
                },
                {
                    "start": 324.64,
                    "end": 324.94,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 324.94,
                    "end": 325.4,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 325.4,
                    "end": 325.98,
                    "word": " grooming",
                    "probability": 0.79150390625
                },
                {
                    "start": 325.98,
                    "end": 326.44,
                    "word": " stage",
                    "probability": 0.74072265625
                },
                {
                    "start": 326.44,
                    "end": 326.72,
                    "word": " of",
                    "probability": 0.8955078125
                },
                {
                    "start": 326.72,
                    "end": 326.82,
                    "word": " an",
                    "probability": 0.90087890625
                },
                {
                    "start": 326.82,
                    "end": 327.22,
                    "word": " attacker's",
                    "probability": 0.863525390625
                },
                {
                    "start": 327.22,
                    "end": 327.58,
                    "word": " workflow",
                    "probability": 0.7509765625
                },
                {
                    "start": 327.58,
                    "end": 327.84,
                    "word": ".",
                    "probability": 0.9208984375
                },
                {
                    "start": 327.84,
                    "end": 328.3,
                    "word": " Here",
                    "probability": 0.8271484375
                },
                {
                    "start": 328.3,
                    "end": 328.78,
                    "word": ",",
                    "probability": 0.66796875
                },
                {
                    "start": 328.78,
                    "end": 328.88,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 328.88,
                    "end": 329.14,
                    "word": " attacker",
                    "probability": 0.71484375
                },
                {
                    "start": 329.14,
                    "end": 329.84,
                    "word": " has",
                    "probability": 0.90087890625
                },
                {
                    "start": 329.84,
                    "end": 330.9,
                    "word": " established",
                    "probability": 0.84423828125
                }
            ]
        },
        {
            "id": 55,
            "seek": 33178,
            "start": 330.9,
            "end": 338.04,
            "text": " themselves as an impersonated chancellor of UC Berkeley. They are emailing an employee at UC",
            "tokens": [
                51132,
                2969,
                382,
                364,
                38147,
                770,
                49225,
                295,
                14079,
                23684,
                13,
                814,
                366,
                3796,
                278,
                364,
                10738,
                412,
                14079,
                51484
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2376838298404918,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0005435943603515625,
            "words": [
                {
                    "start": 330.9,
                    "end": 331.36,
                    "word": " themselves",
                    "probability": 0.6103515625
                },
                {
                    "start": 331.36,
                    "end": 331.66,
                    "word": " as",
                    "probability": 0.9013671875
                },
                {
                    "start": 331.66,
                    "end": 331.8,
                    "word": " an",
                    "probability": 0.90576171875
                },
                {
                    "start": 331.8,
                    "end": 332.52,
                    "word": " impersonated",
                    "probability": 0.88232421875
                },
                {
                    "start": 332.52,
                    "end": 334.52,
                    "word": " chancellor",
                    "probability": 0.59033203125
                },
                {
                    "start": 334.52,
                    "end": 334.76,
                    "word": " of",
                    "probability": 0.890625
                },
                {
                    "start": 334.76,
                    "end": 334.98,
                    "word": " UC",
                    "probability": 0.931640625
                },
                {
                    "start": 334.98,
                    "end": 335.3,
                    "word": " Berkeley",
                    "probability": 0.87548828125
                },
                {
                    "start": 335.3,
                    "end": 335.88,
                    "word": ".",
                    "probability": 0.88232421875
                },
                {
                    "start": 336.0,
                    "end": 336.24,
                    "word": " They",
                    "probability": 0.734375
                },
                {
                    "start": 336.24,
                    "end": 336.56,
                    "word": " are",
                    "probability": 0.89013671875
                },
                {
                    "start": 336.56,
                    "end": 337.1,
                    "word": " emailing",
                    "probability": 0.86865234375
                },
                {
                    "start": 337.1,
                    "end": 337.3,
                    "word": " an",
                    "probability": 0.90478515625
                },
                {
                    "start": 337.3,
                    "end": 337.6,
                    "word": " employee",
                    "probability": 0.6376953125
                },
                {
                    "start": 337.6,
                    "end": 337.8,
                    "word": " at",
                    "probability": 0.90087890625
                },
                {
                    "start": 337.8,
                    "end": 338.04,
                    "word": " UC",
                    "probability": 0.9384765625
                }
            ]
        },
        {
            "id": 56,
            "seek": 33178,
            "start": 338.04,
            "end": 342.64,
            "text": " Berkeley, asking if they're available, looking to exchange messages with them, probably about to",
            "tokens": [
                51484,
                23684,
                11,
                3365,
                498,
                436,
                434,
                2435,
                11,
                1237,
                281,
                7742,
                7897,
                365,
                552,
                11,
                1391,
                466,
                281,
                51712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2376838298404918,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0005435943603515625,
            "words": [
                {
                    "start": 338.04,
                    "end": 338.38,
                    "word": " Berkeley",
                    "probability": 0.90869140625
                },
                {
                    "start": 338.38,
                    "end": 338.6,
                    "word": ",",
                    "probability": 0.60009765625
                },
                {
                    "start": 338.6,
                    "end": 338.78,
                    "word": " asking",
                    "probability": 0.8193359375
                },
                {
                    "start": 338.78,
                    "end": 339.0,
                    "word": " if",
                    "probability": 0.8876953125
                },
                {
                    "start": 339.0,
                    "end": 339.16,
                    "word": " they're",
                    "probability": 0.82373046875
                },
                {
                    "start": 339.16,
                    "end": 339.58,
                    "word": " available",
                    "probability": 0.73974609375
                },
                {
                    "start": 339.58,
                    "end": 340.14,
                    "word": ",",
                    "probability": 0.92041015625
                },
                {
                    "start": 340.14,
                    "end": 340.34,
                    "word": " looking",
                    "probability": 0.79150390625
                },
                {
                    "start": 340.34,
                    "end": 340.5,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 340.5,
                    "end": 340.86,
                    "word": " exchange",
                    "probability": 0.82373046875
                },
                {
                    "start": 340.86,
                    "end": 341.4,
                    "word": " messages",
                    "probability": 0.806640625
                },
                {
                    "start": 341.4,
                    "end": 341.72,
                    "word": " with",
                    "probability": 0.81494140625
                },
                {
                    "start": 341.72,
                    "end": 341.86,
                    "word": " them",
                    "probability": 0.72705078125
                },
                {
                    "start": 341.86,
                    "end": 341.94,
                    "word": ",",
                    "probability": 0.87548828125
                },
                {
                    "start": 341.94,
                    "end": 342.2,
                    "word": " probably",
                    "probability": 0.91064453125
                },
                {
                    "start": 342.2,
                    "end": 342.48,
                    "word": " about",
                    "probability": 0.82373046875
                },
                {
                    "start": 342.48,
                    "end": 342.64,
                    "word": " to",
                    "probability": 0.90625
                }
            ]
        },
        {
            "id": 57,
            "seek": 36068,
            "start": 342.64,
            "end": 351.51,
            "text": " make an ask around a money transfer. Now, I think it's important to highlight why phishing",
            "tokens": [
                50364,
                652,
                364,
                1029,
                926,
                257,
                1460,
                5003,
                13,
                823,
                11,
                286,
                519,
                309,
                311,
                1021,
                281,
                5078,
                983,
                903,
                3807,
                50692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19704860465394125,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.0024318695068359375,
            "words": [
                {
                    "start": 342.64,
                    "end": 342.78,
                    "word": " make",
                    "probability": 0.13134765625
                },
                {
                    "start": 342.78,
                    "end": 342.92,
                    "word": " an",
                    "probability": 0.8369140625
                },
                {
                    "start": 342.92,
                    "end": 343.24,
                    "word": " ask",
                    "probability": 0.79833984375
                },
                {
                    "start": 343.24,
                    "end": 343.72,
                    "word": " around",
                    "probability": 0.83935546875
                },
                {
                    "start": 344.44,
                    "end": 344.62,
                    "word": " a money",
                    "probability": 0.83203125
                },
                {
                    "start": 344.62,
                    "end": 345.06,
                    "word": " transfer",
                    "probability": 0.853515625
                },
                {
                    "start": 345.06,
                    "end": 345.72,
                    "word": ".",
                    "probability": 0.95166015625
                },
                {
                    "start": 348.81,
                    "end": 349.47,
                    "word": " Now",
                    "probability": 0.62353515625
                },
                {
                    "start": 349.47,
                    "end": 349.61,
                    "word": ",",
                    "probability": 0.54296875
                },
                {
                    "start": 349.61,
                    "end": 349.81,
                    "word": " I",
                    "probability": 0.97509765625
                },
                {
                    "start": 349.81,
                    "end": 349.95,
                    "word": " think",
                    "probability": 0.8115234375
                },
                {
                    "start": 349.95,
                    "end": 350.07,
                    "word": " it's",
                    "probability": 0.928955078125
                },
                {
                    "start": 350.07,
                    "end": 350.31,
                    "word": " important",
                    "probability": 0.76953125
                },
                {
                    "start": 350.31,
                    "end": 350.51,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 350.51,
                    "end": 350.75,
                    "word": " highlight",
                    "probability": 0.79296875
                },
                {
                    "start": 350.75,
                    "end": 351.17,
                    "word": " why",
                    "probability": 0.82275390625
                },
                {
                    "start": 351.17,
                    "end": 351.51,
                    "word": " phishing",
                    "probability": 0.888916015625
                }
            ]
        },
        {
            "id": 58,
            "seek": 36068,
            "start": 351.51,
            "end": 359.57,
            "text": " detection is hard. And why in particular, detecting new previously unseen phishing emails written as",
            "tokens": [
                50692,
                17784,
                307,
                1152,
                13,
                400,
                983,
                294,
                1729,
                11,
                40237,
                777,
                8046,
                40608,
                903,
                3807,
                12524,
                3720,
                382,
                51092
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19704860465394125,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.0024318695068359375,
            "words": [
                {
                    "start": 351.51,
                    "end": 351.93,
                    "word": " detection",
                    "probability": 0.74658203125
                },
                {
                    "start": 351.93,
                    "end": 352.35,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 352.35,
                    "end": 352.85,
                    "word": " hard",
                    "probability": 0.810546875
                },
                {
                    "start": 352.85,
                    "end": 353.51,
                    "word": ".",
                    "probability": 0.12066650390625
                },
                {
                    "start": 353.57,
                    "end": 353.67,
                    "word": " And",
                    "probability": 0.7880859375
                },
                {
                    "start": 353.67,
                    "end": 354.97,
                    "word": " why",
                    "probability": 0.74072265625
                },
                {
                    "start": 354.97,
                    "end": 355.11,
                    "word": " in",
                    "probability": 0.56201171875
                },
                {
                    "start": 355.11,
                    "end": 355.49,
                    "word": " particular",
                    "probability": 0.82861328125
                },
                {
                    "start": 355.49,
                    "end": 355.79,
                    "word": ",",
                    "probability": 0.619140625
                },
                {
                    "start": 355.79,
                    "end": 356.17,
                    "word": " detecting",
                    "probability": 0.75732421875
                },
                {
                    "start": 356.17,
                    "end": 356.83,
                    "word": " new",
                    "probability": 0.8232421875
                },
                {
                    "start": 356.83,
                    "end": 357.37,
                    "word": " previously",
                    "probability": 0.410400390625
                },
                {
                    "start": 357.37,
                    "end": 357.85,
                    "word": " unseen",
                    "probability": 0.916015625
                },
                {
                    "start": 357.85,
                    "end": 358.41,
                    "word": " phishing",
                    "probability": 0.89794921875
                },
                {
                    "start": 358.41,
                    "end": 358.79,
                    "word": " emails",
                    "probability": 0.7294921875
                },
                {
                    "start": 358.79,
                    "end": 359.33,
                    "word": " written",
                    "probability": 0.82275390625
                },
                {
                    "start": 359.33,
                    "end": 359.57,
                    "word": " as",
                    "probability": 0.89794921875
                }
            ]
        },
        {
            "id": 59,
            "seek": 36068,
            "start": 359.57,
            "end": 367.59,
            "text": " part of a manual phishing campaign, like what I've just described, is hard. And what this boils down",
            "tokens": [
                51092,
                644,
                295,
                257,
                9688,
                903,
                3807,
                5129,
                11,
                411,
                437,
                286,
                600,
                445,
                7619,
                11,
                307,
                1152,
                13,
                400,
                437,
                341,
                35049,
                760,
                51496
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19704860465394125,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.0024318695068359375,
            "words": [
                {
                    "start": 359.57,
                    "end": 359.81,
                    "word": " part",
                    "probability": 0.802734375
                },
                {
                    "start": 359.81,
                    "end": 360.09,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 360.71,
                    "end": 361.11,
                    "word": " a manual",
                    "probability": 0.91455078125
                },
                {
                    "start": 361.11,
                    "end": 362.01,
                    "word": " phishing",
                    "probability": 0.907470703125
                },
                {
                    "start": 362.01,
                    "end": 362.39,
                    "word": " campaign",
                    "probability": 0.78515625
                },
                {
                    "start": 362.39,
                    "end": 362.49,
                    "word": ",",
                    "probability": 0.33203125
                },
                {
                    "start": 362.49,
                    "end": 362.59,
                    "word": " like",
                    "probability": 0.87158203125
                },
                {
                    "start": 362.59,
                    "end": 362.71,
                    "word": " what",
                    "probability": 0.87353515625
                },
                {
                    "start": 362.71,
                    "end": 362.85,
                    "word": " I've",
                    "probability": 0.94580078125
                },
                {
                    "start": 362.85,
                    "end": 363.09,
                    "word": " just",
                    "probability": 0.8857421875
                },
                {
                    "start": 363.09,
                    "end": 363.73,
                    "word": " described",
                    "probability": 0.81494140625
                },
                {
                    "start": 363.73,
                    "end": 364.09,
                    "word": ",",
                    "probability": 0.76123046875
                },
                {
                    "start": 364.09,
                    "end": 364.55,
                    "word": " is",
                    "probability": 0.9033203125
                },
                {
                    "start": 364.55,
                    "end": 365.07,
                    "word": " hard",
                    "probability": 0.76171875
                },
                {
                    "start": 365.07,
                    "end": 365.73,
                    "word": ".",
                    "probability": 0.99365234375
                },
                {
                    "start": 366.05,
                    "end": 366.67,
                    "word": " And",
                    "probability": 0.8369140625
                },
                {
                    "start": 366.67,
                    "end": 366.87,
                    "word": " what",
                    "probability": 0.87255859375
                },
                {
                    "start": 366.87,
                    "end": 367.03,
                    "word": " this",
                    "probability": 0.88232421875
                },
                {
                    "start": 367.03,
                    "end": 367.31,
                    "word": " boils",
                    "probability": 0.88818359375
                },
                {
                    "start": 367.31,
                    "end": 367.59,
                    "word": " down",
                    "probability": 0.81005859375
                }
            ]
        },
        {
            "id": 60,
            "seek": 36068,
            "start": 367.59,
            "end": 373.87,
            "text": " to, I think, is that classical natural language processing problems are hard. It's hard to",
            "tokens": [
                51496,
                281,
                11,
                286,
                519,
                11,
                307,
                300,
                13735,
                3303,
                2856,
                9007,
                2740,
                366,
                1152,
                13,
                467,
                311,
                1152,
                281,
                51800
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19704860465394125,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.0024318695068359375,
            "words": [
                {
                    "start": 367.59,
                    "end": 367.89,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 367.89,
                    "end": 368.27,
                    "word": ",",
                    "probability": 0.86962890625
                },
                {
                    "start": 368.27,
                    "end": 368.59,
                    "word": " I",
                    "probability": 0.99267578125
                },
                {
                    "start": 368.59,
                    "end": 368.81,
                    "word": " think",
                    "probability": 0.81298828125
                },
                {
                    "start": 368.81,
                    "end": 369.01,
                    "word": ",",
                    "probability": 0.921875
                },
                {
                    "start": 369.01,
                    "end": 369.25,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 369.25,
                    "end": 369.89,
                    "word": " that",
                    "probability": 0.83251953125
                },
                {
                    "start": 369.89,
                    "end": 370.63,
                    "word": " classical",
                    "probability": 0.88134765625
                },
                {
                    "start": 370.63,
                    "end": 371.59,
                    "word": " natural",
                    "probability": 0.759765625
                },
                {
                    "start": 371.59,
                    "end": 371.89,
                    "word": " language",
                    "probability": 0.66748046875
                },
                {
                    "start": 371.89,
                    "end": 372.23,
                    "word": " processing",
                    "probability": 0.78662109375
                },
                {
                    "start": 372.23,
                    "end": 372.63,
                    "word": " problems",
                    "probability": 0.68505859375
                },
                {
                    "start": 372.63,
                    "end": 372.77,
                    "word": " are",
                    "probability": 0.89892578125
                },
                {
                    "start": 372.77,
                    "end": 372.97,
                    "word": " hard",
                    "probability": 0.8193359375
                },
                {
                    "start": 372.97,
                    "end": 373.09,
                    "word": ".",
                    "probability": 0.9775390625
                },
                {
                    "start": 373.09,
                    "end": 373.27,
                    "word": " It's",
                    "probability": 0.94970703125
                },
                {
                    "start": 373.27,
                    "end": 373.47,
                    "word": " hard",
                    "probability": 0.81591796875
                },
                {
                    "start": 373.47,
                    "end": 373.87,
                    "word": " to",
                    "probability": 0.90625
                }
            ]
        },
        {
            "id": 61,
            "seek": 38572,
            "start": 373.87,
            "end": 378.99,
            "text": " get computers to understand language in any meaningful way and reason about language in any",
            "tokens": [
                50380,
                483,
                10807,
                281,
                1223,
                2856,
                294,
                604,
                10995,
                636,
                293,
                1778,
                466,
                2856,
                294,
                604,
                50632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21434294814482713,
            "compression_ratio": 1.7699530516431925,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 373.87,
                    "end": 374.49,
                    "word": " get",
                    "probability": 0.46923828125
                },
                {
                    "start": 374.49,
                    "end": 374.99,
                    "word": " computers",
                    "probability": 0.7529296875
                },
                {
                    "start": 374.99,
                    "end": 375.31,
                    "word": " to",
                    "probability": 0.90771484375
                },
                {
                    "start": 375.31,
                    "end": 375.81,
                    "word": " understand",
                    "probability": 0.74853515625
                },
                {
                    "start": 375.81,
                    "end": 376.19,
                    "word": " language",
                    "probability": 0.671875
                },
                {
                    "start": 376.19,
                    "end": 376.33,
                    "word": " in",
                    "probability": 0.89404296875
                },
                {
                    "start": 376.33,
                    "end": 376.47,
                    "word": " any",
                    "probability": 0.81640625
                },
                {
                    "start": 376.47,
                    "end": 376.81,
                    "word": " meaningful",
                    "probability": 0.86669921875
                },
                {
                    "start": 376.81,
                    "end": 377.51,
                    "word": " way",
                    "probability": 0.88818359375
                },
                {
                    "start": 377.51,
                    "end": 377.89,
                    "word": " and",
                    "probability": 0.69091796875
                },
                {
                    "start": 377.89,
                    "end": 378.17,
                    "word": " reason",
                    "probability": 0.916015625
                },
                {
                    "start": 378.17,
                    "end": 378.41,
                    "word": " about",
                    "probability": 0.8134765625
                },
                {
                    "start": 378.41,
                    "end": 378.71,
                    "word": " language",
                    "probability": 0.671875
                },
                {
                    "start": 378.71,
                    "end": 378.83,
                    "word": " in",
                    "probability": 0.865234375
                },
                {
                    "start": 378.83,
                    "end": 378.99,
                    "word": " any",
                    "probability": 0.82470703125
                }
            ]
        },
        {
            "id": 62,
            "seek": 38572,
            "start": 378.99,
            "end": 386.53,
            "text": " meaningful way. And detecting phishing emails really boils down to building models that have",
            "tokens": [
                50632,
                10995,
                636,
                13,
                400,
                40237,
                903,
                3807,
                12524,
                534,
                35049,
                760,
                281,
                2390,
                5245,
                300,
                362,
                50996
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21434294814482713,
            "compression_ratio": 1.7699530516431925,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 378.99,
                    "end": 379.43,
                    "word": " meaningful",
                    "probability": 0.51318359375
                },
                {
                    "start": 379.43,
                    "end": 379.91,
                    "word": " way",
                    "probability": 0.88134765625
                },
                {
                    "start": 379.91,
                    "end": 380.59,
                    "word": ".",
                    "probability": 0.6416015625
                },
                {
                    "start": 380.59,
                    "end": 380.87,
                    "word": " And",
                    "probability": 0.7783203125
                },
                {
                    "start": 380.87,
                    "end": 381.71,
                    "word": " detecting",
                    "probability": 0.71044921875
                },
                {
                    "start": 381.71,
                    "end": 382.01,
                    "word": " phishing",
                    "probability": 0.886962890625
                },
                {
                    "start": 382.01,
                    "end": 382.31,
                    "word": " emails",
                    "probability": 0.72802734375
                },
                {
                    "start": 382.31,
                    "end": 383.41,
                    "word": " really",
                    "probability": 0.7685546875
                },
                {
                    "start": 383.41,
                    "end": 383.75,
                    "word": " boils",
                    "probability": 0.87939453125
                },
                {
                    "start": 383.75,
                    "end": 384.09,
                    "word": " down",
                    "probability": 0.81494140625
                },
                {
                    "start": 384.09,
                    "end": 384.57,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 384.57,
                    "end": 385.77,
                    "word": " building",
                    "probability": 0.7890625
                },
                {
                    "start": 385.77,
                    "end": 386.13,
                    "word": " models",
                    "probability": 0.7958984375
                },
                {
                    "start": 386.13,
                    "end": 386.33,
                    "word": " that",
                    "probability": 0.880859375
                },
                {
                    "start": 386.33,
                    "end": 386.53,
                    "word": " have",
                    "probability": 0.8818359375
                }
            ]
        },
        {
            "id": 63,
            "seek": 38572,
            "start": 386.53,
            "end": 394.67,
            "text": " some level of understanding of language. So to understand why algorithmically it's hard to make",
            "tokens": [
                50996,
                512,
                1496,
                295,
                3701,
                295,
                2856,
                13,
                407,
                281,
                1223,
                983,
                9284,
                984,
                309,
                311,
                1152,
                281,
                652,
                51400
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21434294814482713,
            "compression_ratio": 1.7699530516431925,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 386.53,
                    "end": 386.91,
                    "word": " some",
                    "probability": 0.8154296875
                },
                {
                    "start": 386.91,
                    "end": 387.45,
                    "word": " level",
                    "probability": 0.82421875
                },
                {
                    "start": 387.45,
                    "end": 387.67,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 387.67,
                    "end": 388.19,
                    "word": " understanding",
                    "probability": 0.72265625
                },
                {
                    "start": 388.19,
                    "end": 388.75,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 388.75,
                    "end": 389.29,
                    "word": " language",
                    "probability": 0.65771484375
                },
                {
                    "start": 389.29,
                    "end": 389.97,
                    "word": ".",
                    "probability": 0.94873046875
                },
                {
                    "start": 390.61,
                    "end": 390.95,
                    "word": " So",
                    "probability": 0.77978515625
                },
                {
                    "start": 390.95,
                    "end": 391.31,
                    "word": " to",
                    "probability": 0.74365234375
                },
                {
                    "start": 391.31,
                    "end": 391.67,
                    "word": " understand",
                    "probability": 0.75830078125
                },
                {
                    "start": 391.67,
                    "end": 392.09,
                    "word": " why",
                    "probability": 0.83203125
                },
                {
                    "start": 392.09,
                    "end": 393.59,
                    "word": " algorithmically",
                    "probability": 0.752197265625
                },
                {
                    "start": 393.59,
                    "end": 393.79,
                    "word": " it's",
                    "probability": 0.88623046875
                },
                {
                    "start": 393.79,
                    "end": 393.99,
                    "word": " hard",
                    "probability": 0.81591796875
                },
                {
                    "start": 393.99,
                    "end": 394.37,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 394.37,
                    "end": 394.67,
                    "word": " make",
                    "probability": 0.87939453125
                }
            ]
        },
        {
            "id": 64,
            "seek": 38572,
            "start": 394.67,
            "end": 398.91,
            "text": " sense of language, let's look at a few classical natural language processing problems. So one of",
            "tokens": [
                51400,
                2020,
                295,
                2856,
                11,
                718,
                311,
                574,
                412,
                257,
                1326,
                13735,
                3303,
                2856,
                9007,
                2740,
                13,
                407,
                472,
                295,
                51608
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21434294814482713,
            "compression_ratio": 1.7699530516431925,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 394.67,
                    "end": 394.93,
                    "word": " sense",
                    "probability": 0.74267578125
                },
                {
                    "start": 394.93,
                    "end": 395.05,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 395.05,
                    "end": 395.41,
                    "word": " language",
                    "probability": 0.66552734375
                },
                {
                    "start": 395.41,
                    "end": 395.57,
                    "word": ",",
                    "probability": 0.80029296875
                },
                {
                    "start": 395.57,
                    "end": 395.75,
                    "word": " let's",
                    "probability": 0.940185546875
                },
                {
                    "start": 395.75,
                    "end": 395.87,
                    "word": " look",
                    "probability": 0.919921875
                },
                {
                    "start": 395.87,
                    "end": 395.95,
                    "word": " at",
                    "probability": 0.9072265625
                },
                {
                    "start": 396.07,
                    "end": 396.17,
                    "word": " a few",
                    "probability": 0.78369140625
                },
                {
                    "start": 396.17,
                    "end": 396.59,
                    "word": " classical",
                    "probability": 0.89501953125
                },
                {
                    "start": 396.59,
                    "end": 396.93,
                    "word": " natural",
                    "probability": 0.72607421875
                },
                {
                    "start": 396.93,
                    "end": 397.19,
                    "word": " language",
                    "probability": 0.65185546875
                },
                {
                    "start": 397.19,
                    "end": 397.51,
                    "word": " processing",
                    "probability": 0.7705078125
                },
                {
                    "start": 397.51,
                    "end": 397.93,
                    "word": " problems",
                    "probability": 0.697265625
                },
                {
                    "start": 397.93,
                    "end": 398.11,
                    "word": ".",
                    "probability": 0.95263671875
                },
                {
                    "start": 398.11,
                    "end": 398.23,
                    "word": " So",
                    "probability": 0.75390625
                },
                {
                    "start": 398.23,
                    "end": 398.79,
                    "word": " one",
                    "probability": 0.74072265625
                },
                {
                    "start": 398.79,
                    "end": 398.91,
                    "word": " of",
                    "probability": 0.904296875
                }
            ]
        },
        {
            "id": 65,
            "seek": 40828,
            "start": 398.91,
            "end": 406.47,
            "text": " those is coreference resolution. To get a sense of what this problem means, consider the following",
            "tokens": [
                50364,
                729,
                307,
                4965,
                5158,
                8669,
                13,
                1407,
                483,
                257,
                2020,
                295,
                437,
                341,
                1154,
                1355,
                11,
                1949,
                264,
                3480,
                50744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21205356747520213,
            "compression_ratio": 1.8439024390243903,
            "no_speech_prob": 0.0240325927734375,
            "words": [
                {
                    "start": 398.91,
                    "end": 399.11,
                    "word": " those",
                    "probability": 0.03790283203125
                },
                {
                    "start": 399.11,
                    "end": 399.29,
                    "word": " is",
                    "probability": 0.81689453125
                },
                {
                    "start": 399.29,
                    "end": 399.81,
                    "word": " coreference",
                    "probability": 0.693359375
                },
                {
                    "start": 399.81,
                    "end": 400.35,
                    "word": " resolution",
                    "probability": 0.6787109375
                },
                {
                    "start": 400.35,
                    "end": 400.75,
                    "word": ".",
                    "probability": 0.51611328125
                },
                {
                    "start": 401.21,
                    "end": 401.61,
                    "word": " To",
                    "probability": 0.873046875
                },
                {
                    "start": 401.61,
                    "end": 402.01,
                    "word": " get",
                    "probability": 0.85986328125
                },
                {
                    "start": 402.13,
                    "end": 402.31,
                    "word": " a sense",
                    "probability": 0.751953125
                },
                {
                    "start": 402.31,
                    "end": 402.55,
                    "word": " of",
                    "probability": 0.90771484375
                },
                {
                    "start": 402.55,
                    "end": 403.17,
                    "word": " what",
                    "probability": 0.8603515625
                },
                {
                    "start": 403.17,
                    "end": 403.37,
                    "word": " this",
                    "probability": 0.88330078125
                },
                {
                    "start": 403.37,
                    "end": 403.71,
                    "word": " problem",
                    "probability": 0.75634765625
                },
                {
                    "start": 403.71,
                    "end": 405.29,
                    "word": " means",
                    "probability": 0.87353515625
                },
                {
                    "start": 405.29,
                    "end": 405.53,
                    "word": ",",
                    "probability": 0.80126953125
                },
                {
                    "start": 405.53,
                    "end": 406.07,
                    "word": " consider",
                    "probability": 0.80712890625
                },
                {
                    "start": 406.07,
                    "end": 406.25,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 406.25,
                    "end": 406.47,
                    "word": " following",
                    "probability": 0.830078125
                }
            ]
        },
        {
            "id": 66,
            "seek": 40828,
            "start": 406.47,
            "end": 411.51,
            "text": " sentence. I went to the store for some milk, and based on the price, decided to buy it.",
            "tokens": [
                50744,
                8174,
                13,
                286,
                1437,
                281,
                264,
                3531,
                337,
                512,
                5392,
                11,
                293,
                2361,
                322,
                264,
                3218,
                11,
                3047,
                281,
                2256,
                309,
                13,
                50968
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21205356747520213,
            "compression_ratio": 1.8439024390243903,
            "no_speech_prob": 0.0240325927734375,
            "words": [
                {
                    "start": 406.47,
                    "end": 406.97,
                    "word": " sentence",
                    "probability": 0.82177734375
                },
                {
                    "start": 406.97,
                    "end": 407.37,
                    "word": ".",
                    "probability": 0.591796875
                },
                {
                    "start": 407.41,
                    "end": 407.49,
                    "word": " I",
                    "probability": 0.703125
                },
                {
                    "start": 407.49,
                    "end": 407.69,
                    "word": " went",
                    "probability": 0.79736328125
                },
                {
                    "start": 407.69,
                    "end": 407.83,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 407.83,
                    "end": 407.97,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 407.97,
                    "end": 408.15,
                    "word": " store",
                    "probability": 0.79150390625
                },
                {
                    "start": 408.15,
                    "end": 408.33,
                    "word": " for",
                    "probability": 0.8955078125
                },
                {
                    "start": 408.33,
                    "end": 408.55,
                    "word": " some",
                    "probability": 0.81787109375
                },
                {
                    "start": 408.55,
                    "end": 408.83,
                    "word": " milk",
                    "probability": 0.8408203125
                },
                {
                    "start": 408.83,
                    "end": 409.07,
                    "word": ",",
                    "probability": 0.302490234375
                },
                {
                    "start": 409.07,
                    "end": 409.17,
                    "word": " and",
                    "probability": 0.8896484375
                },
                {
                    "start": 409.17,
                    "end": 409.37,
                    "word": " based",
                    "probability": 0.79833984375
                },
                {
                    "start": 409.37,
                    "end": 409.51,
                    "word": " on",
                    "probability": 0.90869140625
                },
                {
                    "start": 409.51,
                    "end": 409.63,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 409.63,
                    "end": 409.99,
                    "word": " price",
                    "probability": 0.81982421875
                },
                {
                    "start": 409.99,
                    "end": 410.15,
                    "word": ",",
                    "probability": 0.75830078125
                },
                {
                    "start": 410.15,
                    "end": 410.45,
                    "word": " decided",
                    "probability": 0.853515625
                },
                {
                    "start": 410.45,
                    "end": 410.65,
                    "word": " to",
                    "probability": 0.89892578125
                },
                {
                    "start": 410.65,
                    "end": 410.85,
                    "word": " buy",
                    "probability": 0.81103515625
                },
                {
                    "start": 410.85,
                    "end": 411.11,
                    "word": " it",
                    "probability": 0.90087890625
                },
                {
                    "start": 411.11,
                    "end": 411.51,
                    "word": ".",
                    "probability": 0.9912109375
                }
            ]
        },
        {
            "id": 67,
            "seek": 40828,
            "start": 412.47,
            "end": 417.95,
            "text": " Now, from a grammatical perspective, it could refer to the store or to the milk, right? It's",
            "tokens": [
                51044,
                823,
                11,
                490,
                257,
                17570,
                267,
                804,
                4585,
                11,
                309,
                727,
                2864,
                281,
                264,
                3531,
                420,
                281,
                264,
                5392,
                11,
                558,
                30,
                467,
                311,
                51312
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21205356747520213,
            "compression_ratio": 1.8439024390243903,
            "no_speech_prob": 0.0240325927734375,
            "words": [
                {
                    "start": 412.47,
                    "end": 412.77,
                    "word": " Now",
                    "probability": 0.82666015625
                },
                {
                    "start": 412.77,
                    "end": 412.99,
                    "word": ",",
                    "probability": 0.7138671875
                },
                {
                    "start": 412.99,
                    "end": 413.67,
                    "word": " from",
                    "probability": 0.8173828125
                },
                {
                    "start": 413.81,
                    "end": 414.15,
                    "word": " a grammatical",
                    "probability": 0.8818359375
                },
                {
                    "start": 414.15,
                    "end": 414.57,
                    "word": " perspective",
                    "probability": 0.7890625
                },
                {
                    "start": 414.57,
                    "end": 414.93,
                    "word": ",",
                    "probability": 0.92724609375
                },
                {
                    "start": 414.93,
                    "end": 415.13,
                    "word": " it",
                    "probability": 0.453369140625
                },
                {
                    "start": 415.13,
                    "end": 415.69,
                    "word": " could",
                    "probability": 0.787109375
                },
                {
                    "start": 415.69,
                    "end": 415.91,
                    "word": " refer",
                    "probability": 0.7880859375
                },
                {
                    "start": 415.91,
                    "end": 416.05,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 416.05,
                    "end": 416.23,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 416.23,
                    "end": 416.49,
                    "word": " store",
                    "probability": 0.802734375
                },
                {
                    "start": 416.49,
                    "end": 416.89,
                    "word": " or",
                    "probability": 0.7978515625
                },
                {
                    "start": 416.89,
                    "end": 417.05,
                    "word": " to",
                    "probability": 0.88623046875
                },
                {
                    "start": 417.05,
                    "end": 417.19,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 417.19,
                    "end": 417.41,
                    "word": " milk",
                    "probability": 0.8525390625
                },
                {
                    "start": 417.41,
                    "end": 417.49,
                    "word": ",",
                    "probability": 0.697265625
                },
                {
                    "start": 417.49,
                    "end": 417.63,
                    "word": " right",
                    "probability": 0.861328125
                },
                {
                    "start": 417.63,
                    "end": 417.71,
                    "word": "?",
                    "probability": 0.88916015625
                },
                {
                    "start": 417.71,
                    "end": 417.95,
                    "word": " It's",
                    "probability": 0.88427734375
                }
            ]
        },
        {
            "id": 68,
            "seek": 40828,
            "start": 417.95,
            "end": 421.47,
            "text": " possible that I went to the store for some milk, and based on the price of the store, I decided to",
            "tokens": [
                51312,
                1944,
                300,
                286,
                1437,
                281,
                264,
                3531,
                337,
                512,
                5392,
                11,
                293,
                2361,
                322,
                264,
                3218,
                295,
                264,
                3531,
                11,
                286,
                3047,
                281,
                51488
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21205356747520213,
            "compression_ratio": 1.8439024390243903,
            "no_speech_prob": 0.0240325927734375,
            "words": [
                {
                    "start": 417.95,
                    "end": 418.25,
                    "word": " possible",
                    "probability": 0.8408203125
                },
                {
                    "start": 418.25,
                    "end": 418.41,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 418.41,
                    "end": 418.51,
                    "word": " I",
                    "probability": 0.99609375
                },
                {
                    "start": 418.51,
                    "end": 418.63,
                    "word": " went",
                    "probability": 0.806640625
                },
                {
                    "start": 418.63,
                    "end": 418.73,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 418.73,
                    "end": 418.87,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 418.87,
                    "end": 419.03,
                    "word": " store",
                    "probability": 0.80712890625
                },
                {
                    "start": 419.03,
                    "end": 419.19,
                    "word": " for",
                    "probability": 0.896484375
                },
                {
                    "start": 419.19,
                    "end": 419.39,
                    "word": " some",
                    "probability": 0.8232421875
                },
                {
                    "start": 419.39,
                    "end": 419.57,
                    "word": " milk",
                    "probability": 0.8388671875
                },
                {
                    "start": 419.57,
                    "end": 419.65,
                    "word": ",",
                    "probability": 0.54638671875
                },
                {
                    "start": 419.65,
                    "end": 419.75,
                    "word": " and",
                    "probability": 0.8935546875
                },
                {
                    "start": 419.75,
                    "end": 420.07,
                    "word": " based",
                    "probability": 0.80859375
                },
                {
                    "start": 420.07,
                    "end": 420.17,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 420.17,
                    "end": 420.29,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 420.29,
                    "end": 420.47,
                    "word": " price",
                    "probability": 0.81591796875
                },
                {
                    "start": 420.47,
                    "end": 420.57,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 420.57,
                    "end": 420.69,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 420.69,
                    "end": 420.87,
                    "word": " store",
                    "probability": 0.73046875
                },
                {
                    "start": 420.87,
                    "end": 420.95,
                    "word": ",",
                    "probability": 0.82373046875
                },
                {
                    "start": 420.95,
                    "end": 420.99,
                    "word": " I",
                    "probability": 0.97900390625
                },
                {
                    "start": 420.99,
                    "end": 421.29,
                    "word": " decided",
                    "probability": 0.86669921875
                },
                {
                    "start": 421.29,
                    "end": 421.47,
                    "word": " to",
                    "probability": 0.8974609375
                }
            ]
        },
        {
            "id": 69,
            "seek": 43466,
            "start": 421.47,
            "end": 429.03,
            "text": " buy it. But it's more likely that I bought the milk. As humans, in solving this coreference",
            "tokens": [
                50364,
                2256,
                309,
                13,
                583,
                309,
                311,
                544,
                3700,
                300,
                286,
                4243,
                264,
                5392,
                13,
                1018,
                6255,
                11,
                294,
                12606,
                341,
                4965,
                5158,
                50744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21126995172272336,
            "compression_ratio": 1.6885964912280702,
            "no_speech_prob": 0.0126190185546875,
            "words": [
                {
                    "start": 421.47,
                    "end": 421.61,
                    "word": " buy",
                    "probability": 0.1741943359375
                },
                {
                    "start": 421.61,
                    "end": 421.83,
                    "word": " it",
                    "probability": 0.9013671875
                },
                {
                    "start": 421.83,
                    "end": 422.29,
                    "word": ".",
                    "probability": 0.268310546875
                },
                {
                    "start": 422.41,
                    "end": 422.93,
                    "word": " But",
                    "probability": 0.8173828125
                },
                {
                    "start": 422.93,
                    "end": 423.09,
                    "word": " it's",
                    "probability": 0.943359375
                },
                {
                    "start": 423.09,
                    "end": 423.25,
                    "word": " more",
                    "probability": 0.88427734375
                },
                {
                    "start": 423.25,
                    "end": 423.51,
                    "word": " likely",
                    "probability": 0.802734375
                },
                {
                    "start": 423.51,
                    "end": 423.73,
                    "word": " that",
                    "probability": 0.86865234375
                },
                {
                    "start": 423.73,
                    "end": 424.09,
                    "word": " I",
                    "probability": 0.88232421875
                },
                {
                    "start": 424.09,
                    "end": 424.29,
                    "word": " bought",
                    "probability": 0.69580078125
                },
                {
                    "start": 424.29,
                    "end": 424.47,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 424.47,
                    "end": 424.77,
                    "word": " milk",
                    "probability": 0.849609375
                },
                {
                    "start": 424.77,
                    "end": 425.29,
                    "word": ".",
                    "probability": 0.72998046875
                },
                {
                    "start": 425.49,
                    "end": 425.75,
                    "word": " As",
                    "probability": 0.86279296875
                },
                {
                    "start": 425.75,
                    "end": 426.31,
                    "word": " humans",
                    "probability": 0.84716796875
                },
                {
                    "start": 426.31,
                    "end": 427.13,
                    "word": ",",
                    "probability": 0.603515625
                },
                {
                    "start": 427.13,
                    "end": 427.55,
                    "word": " in",
                    "probability": 0.83203125
                },
                {
                    "start": 427.55,
                    "end": 428.01,
                    "word": " solving",
                    "probability": 0.8037109375
                },
                {
                    "start": 428.01,
                    "end": 428.39,
                    "word": " this",
                    "probability": 0.8876953125
                },
                {
                    "start": 428.39,
                    "end": 429.03,
                    "word": " coreference",
                    "probability": 0.6009521484375
                }
            ]
        },
        {
            "id": 70,
            "seek": 43466,
            "start": 429.03,
            "end": 437.43,
            "text": " resolution problem, so resolving what it refers to, we have to plumb the depths of a number of",
            "tokens": [
                50744,
                8669,
                1154,
                11,
                370,
                49940,
                437,
                309,
                14942,
                281,
                11,
                321,
                362,
                281,
                499,
                2860,
                264,
                28439,
                295,
                257,
                1230,
                295,
                51180
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21126995172272336,
            "compression_ratio": 1.6885964912280702,
            "no_speech_prob": 0.0126190185546875,
            "words": [
                {
                    "start": 429.03,
                    "end": 429.43,
                    "word": " resolution",
                    "probability": 0.6611328125
                },
                {
                    "start": 429.43,
                    "end": 429.79,
                    "word": " problem",
                    "probability": 0.6904296875
                },
                {
                    "start": 429.79,
                    "end": 429.93,
                    "word": ",",
                    "probability": 0.73486328125
                },
                {
                    "start": 429.93,
                    "end": 430.01,
                    "word": " so",
                    "probability": 0.8740234375
                },
                {
                    "start": 430.01,
                    "end": 430.97,
                    "word": " resolving",
                    "probability": 0.70458984375
                },
                {
                    "start": 430.97,
                    "end": 431.39,
                    "word": " what",
                    "probability": 0.88525390625
                },
                {
                    "start": 431.39,
                    "end": 431.61,
                    "word": " it",
                    "probability": 0.86328125
                },
                {
                    "start": 431.61,
                    "end": 431.95,
                    "word": " refers",
                    "probability": 0.798828125
                },
                {
                    "start": 431.95,
                    "end": 432.31,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 432.31,
                    "end": 432.85,
                    "word": ",",
                    "probability": 0.91162109375
                },
                {
                    "start": 432.85,
                    "end": 433.87,
                    "word": " we",
                    "probability": 0.90087890625
                },
                {
                    "start": 433.87,
                    "end": 434.11,
                    "word": " have",
                    "probability": 0.88037109375
                },
                {
                    "start": 434.11,
                    "end": 434.63,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 434.63,
                    "end": 434.97,
                    "word": " plumb",
                    "probability": 0.84912109375
                },
                {
                    "start": 434.97,
                    "end": 435.13,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 435.13,
                    "end": 435.47,
                    "word": " depths",
                    "probability": 0.66455078125
                },
                {
                    "start": 435.47,
                    "end": 436.05,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 436.91,
                    "end": 437.11,
                    "word": " a number",
                    "probability": 0.87451171875
                },
                {
                    "start": 437.11,
                    "end": 437.43,
                    "word": " of",
                    "probability": 0.9091796875
                }
            ]
        },
        {
            "id": 71,
            "seek": 43466,
            "start": 437.43,
            "end": 443.95,
            "text": " complex mental models, right? We need to deploy our syntactical model of the English language,",
            "tokens": [
                51180,
                3997,
                4973,
                5245,
                11,
                558,
                30,
                492,
                643,
                281,
                7274,
                527,
                23980,
                578,
                804,
                2316,
                295,
                264,
                3669,
                2856,
                11,
                51480
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21126995172272336,
            "compression_ratio": 1.6885964912280702,
            "no_speech_prob": 0.0126190185546875,
            "words": [
                {
                    "start": 437.43,
                    "end": 438.25,
                    "word": " complex",
                    "probability": 0.6728515625
                },
                {
                    "start": 438.25,
                    "end": 438.59,
                    "word": " mental",
                    "probability": 0.83837890625
                },
                {
                    "start": 438.59,
                    "end": 438.97,
                    "word": " models",
                    "probability": 0.80126953125
                },
                {
                    "start": 438.97,
                    "end": 439.01,
                    "word": ",",
                    "probability": 0.62060546875
                },
                {
                    "start": 439.01,
                    "end": 439.13,
                    "word": " right",
                    "probability": 0.84765625
                },
                {
                    "start": 439.13,
                    "end": 439.19,
                    "word": "?",
                    "probability": 0.73583984375
                },
                {
                    "start": 439.19,
                    "end": 439.25,
                    "word": " We",
                    "probability": 0.89501953125
                },
                {
                    "start": 439.25,
                    "end": 440.21,
                    "word": " need",
                    "probability": 0.81396484375
                },
                {
                    "start": 440.21,
                    "end": 440.37,
                    "word": " to",
                    "probability": 0.8974609375
                },
                {
                    "start": 440.37,
                    "end": 441.23,
                    "word": " deploy",
                    "probability": 0.6376953125
                },
                {
                    "start": 441.23,
                    "end": 441.63,
                    "word": " our",
                    "probability": 0.81201171875
                },
                {
                    "start": 441.63,
                    "end": 442.21,
                    "word": " syntactical",
                    "probability": 0.8894856770833334
                },
                {
                    "start": 442.21,
                    "end": 442.43,
                    "word": " model",
                    "probability": 0.79248046875
                },
                {
                    "start": 442.43,
                    "end": 442.61,
                    "word": " of",
                    "probability": 0.90478515625
                },
                {
                    "start": 442.61,
                    "end": 442.67,
                    "word": " the",
                    "probability": 0.6396484375
                },
                {
                    "start": 442.67,
                    "end": 442.85,
                    "word": " English",
                    "probability": 0.92529296875
                },
                {
                    "start": 442.85,
                    "end": 443.35,
                    "word": " language",
                    "probability": 0.68017578125
                },
                {
                    "start": 443.35,
                    "end": 443.95,
                    "word": ",",
                    "probability": 0.9501953125
                }
            ]
        },
        {
            "id": 72,
            "seek": 43466,
            "start": 443.95,
            "end": 447.85,
            "text": " our semantic model of the English language, and our model of the world in which a person is more likely",
            "tokens": [
                51480,
                527,
                47982,
                2316,
                295,
                264,
                3669,
                2856,
                11,
                293,
                527,
                2316,
                295,
                264,
                1002,
                294,
                597,
                257,
                954,
                307,
                544,
                3700,
                51708
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21126995172272336,
            "compression_ratio": 1.6885964912280702,
            "no_speech_prob": 0.0126190185546875,
            "words": [
                {
                    "start": 443.95,
                    "end": 444.09,
                    "word": " our",
                    "probability": 0.77587890625
                },
                {
                    "start": 444.09,
                    "end": 444.39,
                    "word": " semantic",
                    "probability": 0.826171875
                },
                {
                    "start": 444.39,
                    "end": 444.69,
                    "word": " model",
                    "probability": 0.79736328125
                },
                {
                    "start": 444.69,
                    "end": 444.83,
                    "word": " of",
                    "probability": 0.89990234375
                },
                {
                    "start": 444.83,
                    "end": 444.89,
                    "word": " the",
                    "probability": 0.6806640625
                },
                {
                    "start": 444.89,
                    "end": 445.05,
                    "word": " English",
                    "probability": 0.9248046875
                },
                {
                    "start": 445.05,
                    "end": 445.33,
                    "word": " language",
                    "probability": 0.6826171875
                },
                {
                    "start": 445.33,
                    "end": 445.35,
                    "word": ",",
                    "probability": 0.568359375
                },
                {
                    "start": 445.35,
                    "end": 445.43,
                    "word": " and",
                    "probability": 0.79541015625
                },
                {
                    "start": 445.43,
                    "end": 445.53,
                    "word": " our",
                    "probability": 0.81640625
                },
                {
                    "start": 445.53,
                    "end": 445.69,
                    "word": " model",
                    "probability": 0.80859375
                },
                {
                    "start": 445.69,
                    "end": 445.83,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 445.83,
                    "end": 445.93,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 445.93,
                    "end": 446.13,
                    "word": " world",
                    "probability": 0.87744140625
                },
                {
                    "start": 446.13,
                    "end": 446.31,
                    "word": " in",
                    "probability": 0.76025390625
                },
                {
                    "start": 446.31,
                    "end": 446.53,
                    "word": " which",
                    "probability": 0.9169921875
                },
                {
                    "start": 446.81,
                    "end": 447.07,
                    "word": " a person",
                    "probability": 0.81201171875
                },
                {
                    "start": 447.07,
                    "end": 447.25,
                    "word": " is",
                    "probability": 0.90576171875
                },
                {
                    "start": 447.25,
                    "end": 447.51,
                    "word": " more",
                    "probability": 0.88330078125
                },
                {
                    "start": 447.51,
                    "end": 447.85,
                    "word": " likely",
                    "probability": 0.7890625
                }
            ]
        },
        {
            "id": 73,
            "seek": 46066,
            "start": 447.85,
            "end": 452.93,
            "text": " to have bought milk at the store than have bought the store itself. And that's how we solve this",
            "tokens": [
                50388,
                281,
                362,
                4243,
                5392,
                412,
                264,
                3531,
                813,
                362,
                4243,
                264,
                3531,
                2564,
                13,
                400,
                300,
                311,
                577,
                321,
                5039,
                341,
                50616
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19687499796802346,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 447.85,
                    "end": 448.51,
                    "word": " to",
                    "probability": 0.54833984375
                },
                {
                    "start": 448.51,
                    "end": 449.05,
                    "word": " have",
                    "probability": 0.86767578125
                },
                {
                    "start": 449.05,
                    "end": 449.55,
                    "word": " bought",
                    "probability": 0.7734375
                },
                {
                    "start": 449.55,
                    "end": 449.81,
                    "word": " milk",
                    "probability": 0.85546875
                },
                {
                    "start": 449.81,
                    "end": 450.01,
                    "word": " at",
                    "probability": 0.90234375
                },
                {
                    "start": 450.01,
                    "end": 450.15,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 450.15,
                    "end": 450.31,
                    "word": " store",
                    "probability": 0.81689453125
                },
                {
                    "start": 450.31,
                    "end": 450.47,
                    "word": " than",
                    "probability": 0.71875
                },
                {
                    "start": 450.47,
                    "end": 450.61,
                    "word": " have",
                    "probability": 0.77197265625
                },
                {
                    "start": 450.61,
                    "end": 450.77,
                    "word": " bought",
                    "probability": 0.78662109375
                },
                {
                    "start": 450.77,
                    "end": 450.93,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 450.93,
                    "end": 451.11,
                    "word": " store",
                    "probability": 0.84521484375
                },
                {
                    "start": 451.11,
                    "end": 451.65,
                    "word": " itself",
                    "probability": 0.68798828125
                },
                {
                    "start": 451.65,
                    "end": 452.09,
                    "word": ".",
                    "probability": 0.4111328125
                },
                {
                    "start": 452.09,
                    "end": 452.17,
                    "word": " And",
                    "probability": 0.78369140625
                },
                {
                    "start": 452.17,
                    "end": 452.35,
                    "word": " that's",
                    "probability": 0.93408203125
                },
                {
                    "start": 452.35,
                    "end": 452.45,
                    "word": " how",
                    "probability": 0.89794921875
                },
                {
                    "start": 452.45,
                    "end": 452.57,
                    "word": " we",
                    "probability": 0.90380859375
                },
                {
                    "start": 452.57,
                    "end": 452.77,
                    "word": " solve",
                    "probability": 0.50537109375
                },
                {
                    "start": 452.77,
                    "end": 452.93,
                    "word": " this",
                    "probability": 0.8125
                }
            ]
        },
        {
            "id": 74,
            "seek": 46066,
            "start": 452.93,
            "end": 459.53,
            "text": " problem. Hopefully, it's clear that it's hard to get algorithms to do that. But to detect phishing",
            "tokens": [
                50616,
                1154,
                13,
                10429,
                11,
                309,
                311,
                1850,
                300,
                309,
                311,
                1152,
                281,
                483,
                14642,
                281,
                360,
                300,
                13,
                583,
                281,
                5531,
                903,
                3807,
                50948
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19687499796802346,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 452.93,
                    "end": 453.19,
                    "word": " problem",
                    "probability": 0.77001953125
                },
                {
                    "start": 453.19,
                    "end": 453.63,
                    "word": ".",
                    "probability": 0.9794921875
                },
                {
                    "start": 454.35,
                    "end": 454.79,
                    "word": " Hopefully",
                    "probability": 0.74658203125
                },
                {
                    "start": 454.79,
                    "end": 454.93,
                    "word": ",",
                    "probability": 0.2369384765625
                },
                {
                    "start": 454.93,
                    "end": 455.01,
                    "word": " it's",
                    "probability": 0.94775390625
                },
                {
                    "start": 455.01,
                    "end": 455.27,
                    "word": " clear",
                    "probability": 0.79443359375
                },
                {
                    "start": 455.27,
                    "end": 456.49,
                    "word": " that",
                    "probability": 0.85693359375
                },
                {
                    "start": 456.49,
                    "end": 456.71,
                    "word": " it's",
                    "probability": 0.94970703125
                },
                {
                    "start": 456.71,
                    "end": 456.83,
                    "word": " hard",
                    "probability": 0.81640625
                },
                {
                    "start": 456.83,
                    "end": 456.93,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 456.93,
                    "end": 457.05,
                    "word": " get",
                    "probability": 0.89794921875
                },
                {
                    "start": 457.05,
                    "end": 457.39,
                    "word": " algorithms",
                    "probability": 0.88671875
                },
                {
                    "start": 457.39,
                    "end": 457.59,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 457.59,
                    "end": 457.69,
                    "word": " do",
                    "probability": 0.91162109375
                },
                {
                    "start": 457.69,
                    "end": 457.93,
                    "word": " that",
                    "probability": 0.88037109375
                },
                {
                    "start": 457.93,
                    "end": 458.37,
                    "word": ".",
                    "probability": 0.7724609375
                },
                {
                    "start": 458.51,
                    "end": 458.89,
                    "word": " But",
                    "probability": 0.884765625
                },
                {
                    "start": 458.89,
                    "end": 459.01,
                    "word": " to",
                    "probability": 0.8759765625
                },
                {
                    "start": 459.01,
                    "end": 459.25,
                    "word": " detect",
                    "probability": 0.892578125
                },
                {
                    "start": 459.25,
                    "end": 459.53,
                    "word": " phishing",
                    "probability": 0.880615234375
                }
            ]
        },
        {
            "id": 75,
            "seek": 46066,
            "start": 459.53,
            "end": 464.89,
            "text": " emails, we really need to understand language. And this is a problem that's constituent of the",
            "tokens": [
                50948,
                12524,
                11,
                321,
                534,
                643,
                281,
                1223,
                2856,
                13,
                400,
                341,
                307,
                257,
                1154,
                300,
                311,
                16085,
                317,
                295,
                264,
                51212
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19687499796802346,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 459.53,
                    "end": 459.83,
                    "word": " emails",
                    "probability": 0.7529296875
                },
                {
                    "start": 459.83,
                    "end": 460.17,
                    "word": ",",
                    "probability": 0.900390625
                },
                {
                    "start": 460.17,
                    "end": 460.49,
                    "word": " we",
                    "probability": 0.90283203125
                },
                {
                    "start": 460.49,
                    "end": 460.67,
                    "word": " really",
                    "probability": 0.7724609375
                },
                {
                    "start": 460.67,
                    "end": 460.81,
                    "word": " need",
                    "probability": 0.81982421875
                },
                {
                    "start": 460.81,
                    "end": 460.89,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 460.89,
                    "end": 461.23,
                    "word": " understand",
                    "probability": 0.7568359375
                },
                {
                    "start": 461.23,
                    "end": 461.67,
                    "word": " language",
                    "probability": 0.66259765625
                },
                {
                    "start": 461.67,
                    "end": 461.87,
                    "word": ".",
                    "probability": 0.69287109375
                },
                {
                    "start": 461.87,
                    "end": 461.99,
                    "word": " And",
                    "probability": 0.79541015625
                },
                {
                    "start": 461.99,
                    "end": 462.19,
                    "word": " this",
                    "probability": 0.87890625
                },
                {
                    "start": 462.19,
                    "end": 462.75,
                    "word": " is",
                    "probability": 0.8876953125
                },
                {
                    "start": 462.99,
                    "end": 463.71,
                    "word": " a problem",
                    "probability": 0.72509765625
                },
                {
                    "start": 463.71,
                    "end": 463.95,
                    "word": " that's",
                    "probability": 0.92236328125
                },
                {
                    "start": 463.95,
                    "end": 464.51,
                    "word": " constituent",
                    "probability": 0.863037109375
                },
                {
                    "start": 464.51,
                    "end": 464.75,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 464.75,
                    "end": 464.89,
                    "word": " the",
                    "probability": 0.81298828125
                }
            ]
        },
        {
            "id": 76,
            "seek": 46066,
            "start": 464.89,
            "end": 469.03,
            "text": " problem of understanding language. Word polysemy is also a classic problem",
            "tokens": [
                51212,
                1154,
                295,
                3701,
                2856,
                13,
                8725,
                6754,
                405,
                2226,
                307,
                611,
                257,
                7230,
                1154,
                51432
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19687499796802346,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 464.89,
                    "end": 465.13,
                    "word": " problem",
                    "probability": 0.76025390625
                },
                {
                    "start": 465.13,
                    "end": 465.27,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 465.27,
                    "end": 465.69,
                    "word": " understanding",
                    "probability": 0.70947265625
                },
                {
                    "start": 465.69,
                    "end": 466.23,
                    "word": " language",
                    "probability": 0.6826171875
                },
                {
                    "start": 466.23,
                    "end": 466.67,
                    "word": ".",
                    "probability": 0.99169921875
                },
                {
                    "start": 467.23,
                    "end": 467.43,
                    "word": " Word",
                    "probability": 0.89111328125
                },
                {
                    "start": 467.43,
                    "end": 467.87,
                    "word": " polysemy",
                    "probability": 0.8258463541666666
                },
                {
                    "start": 467.87,
                    "end": 468.05,
                    "word": " is",
                    "probability": 0.90185546875
                },
                {
                    "start": 468.05,
                    "end": 468.23,
                    "word": " also",
                    "probability": 0.8193359375
                },
                {
                    "start": 468.37,
                    "end": 468.61,
                    "word": " a classic",
                    "probability": 0.8349609375
                },
                {
                    "start": 468.61,
                    "end": 469.03,
                    "word": " problem",
                    "probability": 0.767578125
                }
            ]
        },
        {
            "id": 77,
            "seek": 46066,
            "start": 469.03,
            "end": 473.85,
            "text": " in algorithmic understanding of language. So a sentence like, he drank a lot and was quite the",
            "tokens": [
                51432,
                294,
                9284,
                299,
                3701,
                295,
                2856,
                13,
                407,
                257,
                8174,
                411,
                11,
                415,
                21011,
                257,
                688,
                293,
                390,
                1596,
                264,
                51660
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19687499796802346,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 469.03,
                    "end": 469.49,
                    "word": " in",
                    "probability": 0.85546875
                },
                {
                    "start": 469.49,
                    "end": 470.03,
                    "word": " algorithmic",
                    "probability": 0.854736328125
                },
                {
                    "start": 470.03,
                    "end": 470.51,
                    "word": " understanding",
                    "probability": 0.7236328125
                },
                {
                    "start": 470.51,
                    "end": 470.73,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 470.73,
                    "end": 471.09,
                    "word": " language",
                    "probability": 0.6591796875
                },
                {
                    "start": 471.09,
                    "end": 471.53,
                    "word": ".",
                    "probability": 0.96240234375
                },
                {
                    "start": 471.61,
                    "end": 471.79,
                    "word": " So",
                    "probability": 0.88720703125
                },
                {
                    "start": 471.91,
                    "end": 472.11,
                    "word": " a sentence",
                    "probability": 0.8408203125
                },
                {
                    "start": 472.11,
                    "end": 472.33,
                    "word": " like",
                    "probability": 0.83642578125
                },
                {
                    "start": 472.33,
                    "end": 472.39,
                    "word": ",",
                    "probability": 0.473876953125
                },
                {
                    "start": 472.39,
                    "end": 472.41,
                    "word": " he",
                    "probability": 0.25048828125
                },
                {
                    "start": 472.41,
                    "end": 472.69,
                    "word": " drank",
                    "probability": 0.8330078125
                },
                {
                    "start": 472.85,
                    "end": 473.11,
                    "word": " a lot",
                    "probability": 0.8896484375
                },
                {
                    "start": 473.11,
                    "end": 473.33,
                    "word": " and",
                    "probability": 0.818359375
                },
                {
                    "start": 473.33,
                    "end": 473.47,
                    "word": " was",
                    "probability": 0.89208984375
                },
                {
                    "start": 473.47,
                    "end": 473.67,
                    "word": " quite",
                    "probability": 0.892578125
                },
                {
                    "start": 473.67,
                    "end": 473.85,
                    "word": " the",
                    "probability": 0.8193359375
                }
            ]
        },
        {
            "id": 78,
            "seek": 48778,
            "start": 473.85,
            "end": 480.83,
            "text": " rake. Grammatically, it's valid to interpret this sentence as meaning that he drank a lot",
            "tokens": [
                50364,
                367,
                619,
                13,
                22130,
                76,
                5030,
                11,
                309,
                311,
                7363,
                281,
                7302,
                341,
                8174,
                382,
                3620,
                300,
                415,
                21011,
                257,
                688,
                50712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22906249791383743,
            "compression_ratio": 1.6504424778761062,
            "no_speech_prob": 0.0161285400390625,
            "words": [
                {
                    "start": 473.85,
                    "end": 474.09,
                    "word": " rake",
                    "probability": 0.544891357421875
                },
                {
                    "start": 474.09,
                    "end": 474.55,
                    "word": ".",
                    "probability": 0.56982421875
                },
                {
                    "start": 477.27,
                    "end": 477.73,
                    "word": " Grammatically",
                    "probability": 0.8640950520833334
                },
                {
                    "start": 477.73,
                    "end": 477.97,
                    "word": ",",
                    "probability": 0.53564453125
                },
                {
                    "start": 477.97,
                    "end": 478.09,
                    "word": " it's",
                    "probability": 0.9013671875
                },
                {
                    "start": 478.09,
                    "end": 478.39,
                    "word": " valid",
                    "probability": 0.81787109375
                },
                {
                    "start": 478.39,
                    "end": 478.55,
                    "word": " to",
                    "probability": 0.89599609375
                },
                {
                    "start": 478.55,
                    "end": 478.91,
                    "word": " interpret",
                    "probability": 0.779296875
                },
                {
                    "start": 478.91,
                    "end": 479.29,
                    "word": " this",
                    "probability": 0.62255859375
                },
                {
                    "start": 479.29,
                    "end": 479.57,
                    "word": " sentence",
                    "probability": 0.80859375
                },
                {
                    "start": 479.57,
                    "end": 479.73,
                    "word": " as",
                    "probability": 0.845703125
                },
                {
                    "start": 479.73,
                    "end": 479.95,
                    "word": " meaning",
                    "probability": 0.81884765625
                },
                {
                    "start": 479.95,
                    "end": 480.17,
                    "word": " that",
                    "probability": 0.83642578125
                },
                {
                    "start": 480.17,
                    "end": 480.27,
                    "word": " he",
                    "probability": 0.85888671875
                },
                {
                    "start": 480.27,
                    "end": 480.51,
                    "word": " drank",
                    "probability": 0.8017578125
                },
                {
                    "start": 480.69,
                    "end": 480.83,
                    "word": " a lot",
                    "probability": 0.8955078125
                }
            ]
        },
        {
            "id": 79,
            "seek": 48778,
            "start": 480.83,
            "end": 488.59,
            "text": " and was quite the garden tool used to rake leaves off your lawn. That's clearly not the right,",
            "tokens": [
                50712,
                293,
                390,
                1596,
                264,
                7431,
                2290,
                1143,
                281,
                367,
                619,
                5510,
                766,
                428,
                19915,
                13,
                663,
                311,
                4448,
                406,
                264,
                558,
                11,
                51100
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22906249791383743,
            "compression_ratio": 1.6504424778761062,
            "no_speech_prob": 0.0161285400390625,
            "words": [
                {
                    "start": 480.83,
                    "end": 481.05,
                    "word": " and",
                    "probability": 0.826171875
                },
                {
                    "start": 481.05,
                    "end": 481.25,
                    "word": " was",
                    "probability": 0.87744140625
                },
                {
                    "start": 481.25,
                    "end": 481.79,
                    "word": " quite",
                    "probability": 0.88427734375
                },
                {
                    "start": 481.79,
                    "end": 482.13,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 482.13,
                    "end": 482.97,
                    "word": " garden",
                    "probability": 0.73828125
                },
                {
                    "start": 482.97,
                    "end": 483.35,
                    "word": " tool",
                    "probability": 0.8408203125
                },
                {
                    "start": 483.35,
                    "end": 483.75,
                    "word": " used",
                    "probability": 0.748046875
                },
                {
                    "start": 483.75,
                    "end": 484.59,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 484.59,
                    "end": 485.07,
                    "word": " rake",
                    "probability": 0.896728515625
                },
                {
                    "start": 485.07,
                    "end": 485.35,
                    "word": " leaves",
                    "probability": 0.7099609375
                },
                {
                    "start": 485.35,
                    "end": 485.53,
                    "word": " off",
                    "probability": 0.78759765625
                },
                {
                    "start": 485.53,
                    "end": 485.69,
                    "word": " your",
                    "probability": 0.79296875
                },
                {
                    "start": 485.69,
                    "end": 485.93,
                    "word": " lawn",
                    "probability": 0.79443359375
                },
                {
                    "start": 485.93,
                    "end": 486.39,
                    "word": ".",
                    "probability": 0.98095703125
                },
                {
                    "start": 487.31,
                    "end": 487.61,
                    "word": " That's",
                    "probability": 0.8974609375
                },
                {
                    "start": 487.61,
                    "end": 487.81,
                    "word": " clearly",
                    "probability": 0.81884765625
                },
                {
                    "start": 487.81,
                    "end": 488.09,
                    "word": " not",
                    "probability": 0.8935546875
                },
                {
                    "start": 488.09,
                    "end": 488.23,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 488.23,
                    "end": 488.41,
                    "word": " right",
                    "probability": 0.66552734375
                },
                {
                    "start": 488.41,
                    "end": 488.59,
                    "word": ",",
                    "probability": 0.373779296875
                }
            ]
        },
        {
            "id": 80,
            "seek": 48778,
            "start": 488.59,
            "end": 493.07,
            "text": " that's not the sense in which the word rake is being used. The word rake is being used in the",
            "tokens": [
                51100,
                300,
                311,
                406,
                264,
                2020,
                294,
                597,
                264,
                1349,
                367,
                619,
                307,
                885,
                1143,
                13,
                440,
                1349,
                367,
                619,
                307,
                885,
                1143,
                294,
                264,
                51320
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22906249791383743,
            "compression_ratio": 1.6504424778761062,
            "no_speech_prob": 0.0161285400390625,
            "words": [
                {
                    "start": 488.59,
                    "end": 488.91,
                    "word": " that's",
                    "probability": 0.93017578125
                },
                {
                    "start": 488.91,
                    "end": 489.03,
                    "word": " not",
                    "probability": 0.89306640625
                },
                {
                    "start": 489.03,
                    "end": 489.31,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 489.31,
                    "end": 489.53,
                    "word": " sense",
                    "probability": 0.7373046875
                },
                {
                    "start": 489.53,
                    "end": 489.67,
                    "word": " in",
                    "probability": 0.8388671875
                },
                {
                    "start": 489.67,
                    "end": 489.83,
                    "word": " which",
                    "probability": 0.9130859375
                },
                {
                    "start": 489.83,
                    "end": 489.93,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 489.93,
                    "end": 490.09,
                    "word": " word",
                    "probability": 0.79833984375
                },
                {
                    "start": 490.09,
                    "end": 490.29,
                    "word": " rake",
                    "probability": 0.85400390625
                },
                {
                    "start": 490.29,
                    "end": 490.41,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 490.41,
                    "end": 490.59,
                    "word": " being",
                    "probability": 0.8720703125
                },
                {
                    "start": 490.59,
                    "end": 490.79,
                    "word": " used",
                    "probability": 0.8056640625
                },
                {
                    "start": 490.79,
                    "end": 490.93,
                    "word": ".",
                    "probability": 0.63525390625
                },
                {
                    "start": 490.93,
                    "end": 490.97,
                    "word": " The",
                    "probability": 0.81982421875
                },
                {
                    "start": 490.97,
                    "end": 492.15,
                    "word": " word",
                    "probability": 0.67724609375
                },
                {
                    "start": 492.15,
                    "end": 492.35,
                    "word": " rake",
                    "probability": 0.90869140625
                },
                {
                    "start": 492.35,
                    "end": 492.49,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 492.49,
                    "end": 492.65,
                    "word": " being",
                    "probability": 0.8671875
                },
                {
                    "start": 492.65,
                    "end": 492.85,
                    "word": " used",
                    "probability": 0.7861328125
                },
                {
                    "start": 492.85,
                    "end": 492.95,
                    "word": " in",
                    "probability": 0.8935546875
                },
                {
                    "start": 492.95,
                    "end": 493.07,
                    "word": " the",
                    "probability": 0.7646484375
                }
            ]
        },
        {
            "id": 81,
            "seek": 48778,
            "start": 493.07,
            "end": 500.97,
            "text": " sense of a drunk, semi-criminal, sort of dissolute individual here. But it takes a pretty deep",
            "tokens": [
                51320,
                2020,
                295,
                257,
                11192,
                11,
                12909,
                12,
                66,
                46885,
                11,
                1333,
                295,
                15840,
                1169,
                2609,
                510,
                13,
                583,
                309,
                2516,
                257,
                1238,
                2452,
                51720
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22906249791383743,
            "compression_ratio": 1.6504424778761062,
            "no_speech_prob": 0.0161285400390625,
            "words": [
                {
                    "start": 493.07,
                    "end": 493.25,
                    "word": " sense",
                    "probability": 0.7109375
                },
                {
                    "start": 493.25,
                    "end": 493.41,
                    "word": " of",
                    "probability": 0.8974609375
                },
                {
                    "start": 493.73,
                    "end": 494.03,
                    "word": " a drunk",
                    "probability": 0.76953125
                },
                {
                    "start": 494.03,
                    "end": 494.23,
                    "word": ",",
                    "probability": 0.736328125
                },
                {
                    "start": 494.23,
                    "end": 494.57,
                    "word": " semi",
                    "probability": 0.67724609375
                },
                {
                    "start": 494.57,
                    "end": 494.91,
                    "word": "-criminal",
                    "probability": 0.8592122395833334
                },
                {
                    "start": 494.91,
                    "end": 495.27,
                    "word": ",",
                    "probability": 0.8359375
                },
                {
                    "start": 495.27,
                    "end": 495.51,
                    "word": " sort",
                    "probability": 0.59521484375
                },
                {
                    "start": 495.51,
                    "end": 495.63,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 495.63,
                    "end": 496.15,
                    "word": " dissolute",
                    "probability": 0.84912109375
                },
                {
                    "start": 496.15,
                    "end": 497.25,
                    "word": " individual",
                    "probability": 0.7353515625
                },
                {
                    "start": 497.25,
                    "end": 497.89,
                    "word": " here",
                    "probability": 0.77734375
                },
                {
                    "start": 497.89,
                    "end": 498.35,
                    "word": ".",
                    "probability": 0.69140625
                },
                {
                    "start": 498.57,
                    "end": 499.03,
                    "word": " But",
                    "probability": 0.77490234375
                },
                {
                    "start": 499.03,
                    "end": 499.27,
                    "word": " it",
                    "probability": 0.8447265625
                },
                {
                    "start": 499.27,
                    "end": 499.59,
                    "word": " takes",
                    "probability": 0.71826171875
                },
                {
                    "start": 500.47,
                    "end": 500.63,
                    "word": " a pretty",
                    "probability": 0.8984375
                },
                {
                    "start": 500.63,
                    "end": 500.97,
                    "word": " deep",
                    "probability": 0.9052734375
                }
            ]
        },
        {
            "id": 82,
            "seek": 51594,
            "start": 500.97,
            "end": 507.87,
            "text": " exercise of a human being's mental models to arrive at the sense in which this word was used.",
            "tokens": [
                50364,
                5380,
                295,
                257,
                1952,
                885,
                311,
                4973,
                5245,
                281,
                8881,
                412,
                264,
                2020,
                294,
                597,
                341,
                1349,
                390,
                1143,
                13,
                50700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23008848924552444,
            "compression_ratio": 1.6206896551724137,
            "no_speech_prob": 0.00211334228515625,
            "words": [
                {
                    "start": 500.97,
                    "end": 501.35,
                    "word": " exercise",
                    "probability": 0.376708984375
                },
                {
                    "start": 501.35,
                    "end": 502.03,
                    "word": " of",
                    "probability": 0.8779296875
                },
                {
                    "start": 503.29,
                    "end": 503.49,
                    "word": " a human",
                    "probability": 0.80517578125
                },
                {
                    "start": 503.49,
                    "end": 503.99,
                    "word": " being's",
                    "probability": 0.837158203125
                },
                {
                    "start": 503.99,
                    "end": 504.37,
                    "word": " mental",
                    "probability": 0.81640625
                },
                {
                    "start": 504.37,
                    "end": 504.79,
                    "word": " models",
                    "probability": 0.78564453125
                },
                {
                    "start": 504.79,
                    "end": 505.09,
                    "word": " to",
                    "probability": 0.88427734375
                },
                {
                    "start": 505.09,
                    "end": 505.81,
                    "word": " arrive",
                    "probability": 0.501953125
                },
                {
                    "start": 505.81,
                    "end": 506.07,
                    "word": " at",
                    "probability": 0.8935546875
                },
                {
                    "start": 506.07,
                    "end": 506.53,
                    "word": " the",
                    "probability": 0.77001953125
                },
                {
                    "start": 506.53,
                    "end": 506.75,
                    "word": " sense",
                    "probability": 0.689453125
                },
                {
                    "start": 506.75,
                    "end": 506.87,
                    "word": " in",
                    "probability": 0.88134765625
                },
                {
                    "start": 506.87,
                    "end": 506.99,
                    "word": " which",
                    "probability": 0.90869140625
                },
                {
                    "start": 506.99,
                    "end": 507.17,
                    "word": " this",
                    "probability": 0.873046875
                },
                {
                    "start": 507.17,
                    "end": 507.33,
                    "word": " word",
                    "probability": 0.783203125
                },
                {
                    "start": 507.33,
                    "end": 507.49,
                    "word": " was",
                    "probability": 0.880859375
                },
                {
                    "start": 507.49,
                    "end": 507.71,
                    "word": " used",
                    "probability": 0.794921875
                },
                {
                    "start": 507.71,
                    "end": 507.87,
                    "word": ".",
                    "probability": 0.125732421875
                }
            ]
        },
        {
            "id": 83,
            "seek": 51594,
            "start": 507.87,
            "end": 513.15,
            "text": " And not easy to reproduce in the form of an automated agent, either machine learning or",
            "tokens": [
                50700,
                400,
                406,
                1858,
                281,
                29501,
                294,
                264,
                1254,
                295,
                364,
                18473,
                9461,
                11,
                2139,
                3479,
                2539,
                420,
                50992
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23008848924552444,
            "compression_ratio": 1.6206896551724137,
            "no_speech_prob": 0.00211334228515625,
            "words": [
                {
                    "start": 507.87,
                    "end": 507.91,
                    "word": " And",
                    "probability": 0.64208984375
                },
                {
                    "start": 507.91,
                    "end": 508.69,
                    "word": " not",
                    "probability": 0.82080078125
                },
                {
                    "start": 508.69,
                    "end": 508.91,
                    "word": " easy",
                    "probability": 0.8173828125
                },
                {
                    "start": 508.91,
                    "end": 509.05,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 509.05,
                    "end": 509.37,
                    "word": " reproduce",
                    "probability": 0.81396484375
                },
                {
                    "start": 509.37,
                    "end": 510.63,
                    "word": " in",
                    "probability": 0.82958984375
                },
                {
                    "start": 510.63,
                    "end": 510.89,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 510.89,
                    "end": 511.01,
                    "word": " form",
                    "probability": 0.81201171875
                },
                {
                    "start": 511.01,
                    "end": 511.11,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 511.11,
                    "end": 511.23,
                    "word": " an",
                    "probability": 0.89990234375
                },
                {
                    "start": 511.23,
                    "end": 511.53,
                    "word": " automated",
                    "probability": 0.87451171875
                },
                {
                    "start": 511.53,
                    "end": 511.95,
                    "word": " agent",
                    "probability": 0.76416015625
                },
                {
                    "start": 511.95,
                    "end": 512.07,
                    "word": ",",
                    "probability": 0.55859375
                },
                {
                    "start": 512.07,
                    "end": 512.17,
                    "word": " either",
                    "probability": 0.7744140625
                },
                {
                    "start": 512.17,
                    "end": 512.47,
                    "word": " machine",
                    "probability": 0.66845703125
                },
                {
                    "start": 512.47,
                    "end": 512.73,
                    "word": " learning",
                    "probability": 0.81982421875
                },
                {
                    "start": 512.73,
                    "end": 513.15,
                    "word": " or",
                    "probability": 0.7998046875
                }
            ]
        },
        {
            "id": 84,
            "seek": 51594,
            "start": 513.15,
            "end": 518.43,
            "text": " based on regexes and rules. Sentiment detection is another hard problem in natural language",
            "tokens": [
                50992,
                2361,
                322,
                319,
                432,
                47047,
                293,
                4474,
                13,
                23652,
                2328,
                17784,
                307,
                1071,
                1152,
                1154,
                294,
                3303,
                2856,
                51236
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23008848924552444,
            "compression_ratio": 1.6206896551724137,
            "no_speech_prob": 0.00211334228515625,
            "words": [
                {
                    "start": 513.15,
                    "end": 513.81,
                    "word": " based",
                    "probability": 0.73681640625
                },
                {
                    "start": 513.81,
                    "end": 513.97,
                    "word": " on",
                    "probability": 0.900390625
                },
                {
                    "start": 513.97,
                    "end": 514.33,
                    "word": " regexes",
                    "probability": 0.7421875
                },
                {
                    "start": 514.33,
                    "end": 514.59,
                    "word": " and",
                    "probability": 0.890625
                },
                {
                    "start": 514.59,
                    "end": 514.89,
                    "word": " rules",
                    "probability": 0.73046875
                },
                {
                    "start": 514.89,
                    "end": 515.33,
                    "word": ".",
                    "probability": 0.98974609375
                },
                {
                    "start": 515.99,
                    "end": 516.29,
                    "word": " Sentiment",
                    "probability": 0.81005859375
                },
                {
                    "start": 516.29,
                    "end": 516.63,
                    "word": " detection",
                    "probability": 0.7353515625
                },
                {
                    "start": 516.63,
                    "end": 516.83,
                    "word": " is",
                    "probability": 0.8974609375
                },
                {
                    "start": 516.83,
                    "end": 517.11,
                    "word": " another",
                    "probability": 0.82666015625
                },
                {
                    "start": 517.11,
                    "end": 517.57,
                    "word": " hard",
                    "probability": 0.775390625
                },
                {
                    "start": 517.57,
                    "end": 517.87,
                    "word": " problem",
                    "probability": 0.77587890625
                },
                {
                    "start": 517.87,
                    "end": 517.97,
                    "word": " in",
                    "probability": 0.8564453125
                },
                {
                    "start": 517.97,
                    "end": 518.19,
                    "word": " natural",
                    "probability": 0.74072265625
                },
                {
                    "start": 518.19,
                    "end": 518.43,
                    "word": " language",
                    "probability": 0.673828125
                }
            ]
        },
        {
            "id": 85,
            "seek": 51594,
            "start": 518.43,
            "end": 523.45,
            "text": " processing. So consider the sentence, I'm not angry at all. No, of course. Why would I be angry",
            "tokens": [
                51236,
                9007,
                13,
                407,
                1949,
                264,
                8174,
                11,
                286,
                478,
                406,
                6884,
                412,
                439,
                13,
                883,
                11,
                295,
                1164,
                13,
                1545,
                576,
                286,
                312,
                6884,
                51484
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23008848924552444,
            "compression_ratio": 1.6206896551724137,
            "no_speech_prob": 0.00211334228515625,
            "words": [
                {
                    "start": 518.43,
                    "end": 518.85,
                    "word": " processing",
                    "probability": 0.806640625
                },
                {
                    "start": 518.85,
                    "end": 519.29,
                    "word": ".",
                    "probability": 0.94140625
                },
                {
                    "start": 519.95,
                    "end": 520.09,
                    "word": " So",
                    "probability": 0.75439453125
                },
                {
                    "start": 520.09,
                    "end": 520.49,
                    "word": " consider",
                    "probability": 0.64111328125
                },
                {
                    "start": 520.49,
                    "end": 520.65,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 520.65,
                    "end": 520.97,
                    "word": " sentence",
                    "probability": 0.83447265625
                },
                {
                    "start": 520.97,
                    "end": 521.07,
                    "word": ",",
                    "probability": 0.662109375
                },
                {
                    "start": 521.07,
                    "end": 521.21,
                    "word": " I'm",
                    "probability": 0.6351318359375
                },
                {
                    "start": 521.21,
                    "end": 521.35,
                    "word": " not",
                    "probability": 0.89990234375
                },
                {
                    "start": 521.35,
                    "end": 521.57,
                    "word": " angry",
                    "probability": 0.76806640625
                },
                {
                    "start": 521.57,
                    "end": 521.71,
                    "word": " at",
                    "probability": 0.89990234375
                },
                {
                    "start": 521.71,
                    "end": 521.89,
                    "word": " all",
                    "probability": 0.8984375
                },
                {
                    "start": 521.89,
                    "end": 522.01,
                    "word": ".",
                    "probability": 0.83837890625
                },
                {
                    "start": 522.01,
                    "end": 522.15,
                    "word": " No",
                    "probability": 0.92919921875
                },
                {
                    "start": 522.15,
                    "end": 522.21,
                    "word": ",",
                    "probability": 0.8662109375
                },
                {
                    "start": 522.21,
                    "end": 522.35,
                    "word": " of",
                    "probability": 0.90869140625
                },
                {
                    "start": 522.35,
                    "end": 522.59,
                    "word": " course",
                    "probability": 0.8701171875
                },
                {
                    "start": 522.59,
                    "end": 522.77,
                    "word": ".",
                    "probability": 0.65087890625
                },
                {
                    "start": 522.77,
                    "end": 522.87,
                    "word": " Why",
                    "probability": 0.796875
                },
                {
                    "start": 522.87,
                    "end": 523.01,
                    "word": " would",
                    "probability": 0.8505859375
                },
                {
                    "start": 523.01,
                    "end": 523.09,
                    "word": " I",
                    "probability": 0.99853515625
                },
                {
                    "start": 523.09,
                    "end": 523.25,
                    "word": " be",
                    "probability": 0.90087890625
                },
                {
                    "start": 523.25,
                    "end": 523.45,
                    "word": " angry",
                    "probability": 0.76904296875
                }
            ]
        },
        {
            "id": 86,
            "seek": 51594,
            "start": 523.45,
            "end": 529.13,
            "text": " that you spent our life savings on your mistress? So clearly, the speaker is angry here. And they're",
            "tokens": [
                51484,
                300,
                291,
                4418,
                527,
                993,
                13454,
                322,
                428,
                46635,
                30,
                407,
                4448,
                11,
                264,
                8145,
                307,
                6884,
                510,
                13,
                400,
                436,
                434,
                51764
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23008848924552444,
            "compression_ratio": 1.6206896551724137,
            "no_speech_prob": 0.00211334228515625,
            "words": [
                {
                    "start": 523.45,
                    "end": 523.61,
                    "word": " that",
                    "probability": 0.8818359375
                },
                {
                    "start": 523.61,
                    "end": 523.71,
                    "word": " you",
                    "probability": 0.89599609375
                },
                {
                    "start": 523.71,
                    "end": 523.93,
                    "word": " spent",
                    "probability": 0.80078125
                },
                {
                    "start": 523.93,
                    "end": 524.07,
                    "word": " our",
                    "probability": 0.802734375
                },
                {
                    "start": 524.07,
                    "end": 524.29,
                    "word": " life",
                    "probability": 0.8017578125
                },
                {
                    "start": 524.29,
                    "end": 524.59,
                    "word": " savings",
                    "probability": 0.685546875
                },
                {
                    "start": 524.59,
                    "end": 524.77,
                    "word": " on",
                    "probability": 0.90625
                },
                {
                    "start": 524.77,
                    "end": 524.91,
                    "word": " your",
                    "probability": 0.81103515625
                },
                {
                    "start": 524.91,
                    "end": 525.15,
                    "word": " mistress",
                    "probability": 0.896484375
                },
                {
                    "start": 525.15,
                    "end": 525.59,
                    "word": "?",
                    "probability": 0.94970703125
                },
                {
                    "start": 526.59,
                    "end": 526.83,
                    "word": " So",
                    "probability": 0.880859375
                },
                {
                    "start": 526.83,
                    "end": 527.21,
                    "word": " clearly",
                    "probability": 0.7890625
                },
                {
                    "start": 527.21,
                    "end": 527.35,
                    "word": ",",
                    "probability": 0.313720703125
                },
                {
                    "start": 527.35,
                    "end": 527.51,
                    "word": " the",
                    "probability": 0.6494140625
                },
                {
                    "start": 527.51,
                    "end": 527.93,
                    "word": " speaker",
                    "probability": 0.79248046875
                },
                {
                    "start": 527.93,
                    "end": 528.13,
                    "word": " is",
                    "probability": 0.90869140625
                },
                {
                    "start": 528.13,
                    "end": 528.33,
                    "word": " angry",
                    "probability": 0.7939453125
                },
                {
                    "start": 528.33,
                    "end": 528.61,
                    "word": " here",
                    "probability": 0.81201171875
                },
                {
                    "start": 528.61,
                    "end": 528.81,
                    "word": ".",
                    "probability": 0.2254638671875
                },
                {
                    "start": 528.81,
                    "end": 529.01,
                    "word": " And",
                    "probability": 0.88720703125
                },
                {
                    "start": 529.01,
                    "end": 529.13,
                    "word": " they're",
                    "probability": 0.83349609375
                }
            ]
        },
        {
            "id": 87,
            "seek": 54186,
            "start": 529.13,
            "end": 534.63,
            "text": " being sarcastic. Detecting that they're angry and sarcastic is not trivial and requires that",
            "tokens": [
                50364,
                885,
                36836,
                2750,
                13,
                4237,
                557,
                278,
                300,
                436,
                434,
                6884,
                293,
                36836,
                2750,
                307,
                406,
                26703,
                293,
                7029,
                300,
                50636
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2280159849067067,
            "compression_ratio": 1.7464788732394365,
            "no_speech_prob": 0.0013446807861328125,
            "words": [
                {
                    "start": 529.13,
                    "end": 529.25,
                    "word": " being",
                    "probability": 0.17138671875
                },
                {
                    "start": 529.25,
                    "end": 529.89,
                    "word": " sarcastic",
                    "probability": 0.857421875
                },
                {
                    "start": 529.89,
                    "end": 530.37,
                    "word": ".",
                    "probability": 0.0802001953125
                },
                {
                    "start": 530.37,
                    "end": 530.79,
                    "word": " Detecting",
                    "probability": 0.9064127604166666
                },
                {
                    "start": 530.79,
                    "end": 530.93,
                    "word": " that",
                    "probability": 0.84765625
                },
                {
                    "start": 530.93,
                    "end": 531.05,
                    "word": " they're",
                    "probability": 0.792724609375
                },
                {
                    "start": 531.05,
                    "end": 531.29,
                    "word": " angry",
                    "probability": 0.7880859375
                },
                {
                    "start": 531.29,
                    "end": 531.53,
                    "word": " and",
                    "probability": 0.8798828125
                },
                {
                    "start": 531.53,
                    "end": 532.15,
                    "word": " sarcastic",
                    "probability": 0.844482421875
                },
                {
                    "start": 532.15,
                    "end": 533.19,
                    "word": " is",
                    "probability": 0.87890625
                },
                {
                    "start": 533.19,
                    "end": 533.67,
                    "word": " not",
                    "probability": 0.888671875
                },
                {
                    "start": 533.67,
                    "end": 533.99,
                    "word": " trivial",
                    "probability": 0.841796875
                },
                {
                    "start": 533.99,
                    "end": 534.17,
                    "word": " and",
                    "probability": 0.76171875
                },
                {
                    "start": 534.17,
                    "end": 534.45,
                    "word": " requires",
                    "probability": 0.72607421875
                },
                {
                    "start": 534.45,
                    "end": 534.63,
                    "word": " that",
                    "probability": 0.8701171875
                }
            ]
        },
        {
            "id": 88,
            "seek": 54186,
            "start": 534.63,
            "end": 540.01,
            "text": " we understand not only the syntactic structure of the sentence, but also the semantics of the",
            "tokens": [
                50636,
                321,
                1223,
                406,
                787,
                264,
                23980,
                19892,
                3877,
                295,
                264,
                8174,
                11,
                457,
                611,
                264,
                4361,
                45298,
                295,
                264,
                50904
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2280159849067067,
            "compression_ratio": 1.7464788732394365,
            "no_speech_prob": 0.0013446807861328125,
            "words": [
                {
                    "start": 534.63,
                    "end": 534.75,
                    "word": " we",
                    "probability": 0.908203125
                },
                {
                    "start": 534.75,
                    "end": 535.35,
                    "word": " understand",
                    "probability": 0.74853515625
                },
                {
                    "start": 535.35,
                    "end": 536.91,
                    "word": " not",
                    "probability": 0.79736328125
                },
                {
                    "start": 536.91,
                    "end": 537.13,
                    "word": " only",
                    "probability": 0.85400390625
                },
                {
                    "start": 537.13,
                    "end": 537.29,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 537.29,
                    "end": 537.65,
                    "word": " syntactic",
                    "probability": 0.88916015625
                },
                {
                    "start": 537.65,
                    "end": 537.99,
                    "word": " structure",
                    "probability": 0.69482421875
                },
                {
                    "start": 537.99,
                    "end": 538.15,
                    "word": " of",
                    "probability": 0.89599609375
                },
                {
                    "start": 538.15,
                    "end": 538.25,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 538.25,
                    "end": 538.49,
                    "word": " sentence",
                    "probability": 0.83984375
                },
                {
                    "start": 538.49,
                    "end": 538.61,
                    "word": ",",
                    "probability": 0.39453125
                },
                {
                    "start": 538.61,
                    "end": 538.69,
                    "word": " but",
                    "probability": 0.89306640625
                },
                {
                    "start": 538.69,
                    "end": 538.85,
                    "word": " also",
                    "probability": 0.833984375
                },
                {
                    "start": 538.85,
                    "end": 539.09,
                    "word": " the",
                    "probability": 0.783203125
                },
                {
                    "start": 539.09,
                    "end": 539.67,
                    "word": " semantics",
                    "probability": 0.941162109375
                },
                {
                    "start": 539.67,
                    "end": 539.87,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 539.87,
                    "end": 540.01,
                    "word": " the",
                    "probability": 0.8125
                }
            ]
        },
        {
            "id": 89,
            "seek": 54186,
            "start": 540.01,
            "end": 545.47,
            "text": " sentence. And requires that we have the reflex that this person is probably angry if their",
            "tokens": [
                50904,
                8174,
                13,
                400,
                7029,
                300,
                321,
                362,
                264,
                23802,
                300,
                341,
                954,
                307,
                1391,
                6884,
                498,
                641,
                51180
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2280159849067067,
            "compression_ratio": 1.7464788732394365,
            "no_speech_prob": 0.0013446807861328125,
            "words": [
                {
                    "start": 540.01,
                    "end": 540.41,
                    "word": " sentence",
                    "probability": 0.82177734375
                },
                {
                    "start": 540.41,
                    "end": 540.93,
                    "word": ".",
                    "probability": 0.309814453125
                },
                {
                    "start": 540.93,
                    "end": 541.33,
                    "word": " And",
                    "probability": 0.367919921875
                },
                {
                    "start": 541.33,
                    "end": 541.73,
                    "word": " requires",
                    "probability": 0.33349609375
                },
                {
                    "start": 541.73,
                    "end": 541.93,
                    "word": " that",
                    "probability": 0.87548828125
                },
                {
                    "start": 541.93,
                    "end": 542.01,
                    "word": " we",
                    "probability": 0.90869140625
                },
                {
                    "start": 542.01,
                    "end": 542.15,
                    "word": " have",
                    "probability": 0.8837890625
                },
                {
                    "start": 542.15,
                    "end": 542.27,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 542.27,
                    "end": 542.61,
                    "word": " reflex",
                    "probability": 0.83935546875
                },
                {
                    "start": 542.61,
                    "end": 542.95,
                    "word": " that",
                    "probability": 0.86279296875
                },
                {
                    "start": 542.95,
                    "end": 543.65,
                    "word": " this",
                    "probability": 0.6455078125
                },
                {
                    "start": 543.65,
                    "end": 543.91,
                    "word": " person",
                    "probability": 0.7958984375
                },
                {
                    "start": 543.91,
                    "end": 544.01,
                    "word": " is",
                    "probability": 0.88623046875
                },
                {
                    "start": 544.01,
                    "end": 544.29,
                    "word": " probably",
                    "probability": 0.89892578125
                },
                {
                    "start": 544.29,
                    "end": 544.63,
                    "word": " angry",
                    "probability": 0.7841796875
                },
                {
                    "start": 544.63,
                    "end": 544.93,
                    "word": " if",
                    "probability": 0.85888671875
                },
                {
                    "start": 544.93,
                    "end": 545.47,
                    "word": " their",
                    "probability": 0.74365234375
                }
            ]
        },
        {
            "id": 90,
            "seek": 54186,
            "start": 545.47,
            "end": 555.05,
            "text": " interlocutor spent their life savings on their mistress. Okay, so to solve phishing means that",
            "tokens": [
                51180,
                728,
                5842,
                22163,
                4418,
                641,
                993,
                13454,
                322,
                641,
                46635,
                13,
                1033,
                11,
                370,
                281,
                5039,
                903,
                3807,
                1355,
                300,
                51656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2280159849067067,
            "compression_ratio": 1.7464788732394365,
            "no_speech_prob": 0.0013446807861328125,
            "words": [
                {
                    "start": 545.47,
                    "end": 546.15,
                    "word": " interlocutor",
                    "probability": 0.8815104166666666
                },
                {
                    "start": 546.15,
                    "end": 547.65,
                    "word": " spent",
                    "probability": 0.78125
                },
                {
                    "start": 547.65,
                    "end": 548.35,
                    "word": " their",
                    "probability": 0.89990234375
                },
                {
                    "start": 548.35,
                    "end": 548.59,
                    "word": " life",
                    "probability": 0.7744140625
                },
                {
                    "start": 548.59,
                    "end": 548.91,
                    "word": " savings",
                    "probability": 0.693359375
                },
                {
                    "start": 548.91,
                    "end": 549.21,
                    "word": " on",
                    "probability": 0.90625
                },
                {
                    "start": 549.21,
                    "end": 549.45,
                    "word": " their",
                    "probability": 0.90283203125
                },
                {
                    "start": 549.45,
                    "end": 549.81,
                    "word": " mistress",
                    "probability": 0.8994140625
                },
                {
                    "start": 549.81,
                    "end": 550.41,
                    "word": ".",
                    "probability": 0.98828125
                },
                {
                    "start": 551.29,
                    "end": 551.89,
                    "word": " Okay",
                    "probability": 0.45703125
                },
                {
                    "start": 551.89,
                    "end": 551.99,
                    "word": ",",
                    "probability": 0.71826171875
                },
                {
                    "start": 551.99,
                    "end": 552.11,
                    "word": " so",
                    "probability": 0.8837890625
                },
                {
                    "start": 552.11,
                    "end": 552.87,
                    "word": " to",
                    "probability": 0.81982421875
                },
                {
                    "start": 552.87,
                    "end": 553.19,
                    "word": " solve",
                    "probability": 0.8427734375
                },
                {
                    "start": 553.19,
                    "end": 554.19,
                    "word": " phishing",
                    "probability": 0.820068359375
                },
                {
                    "start": 554.19,
                    "end": 554.85,
                    "word": " means",
                    "probability": 0.84033203125
                },
                {
                    "start": 554.85,
                    "end": 555.05,
                    "word": " that",
                    "probability": 0.8720703125
                }
            ]
        },
        {
            "id": 91,
            "seek": 56922,
            "start": 555.05,
            "end": 559.33,
            "text": " we need algorithms that can make sense of language. Making sense of language is hard, as these three",
            "tokens": [
                50364,
                321,
                643,
                14642,
                300,
                393,
                652,
                2020,
                295,
                2856,
                13,
                14595,
                2020,
                295,
                2856,
                307,
                1152,
                11,
                382,
                613,
                1045,
                50580
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26487379521131516,
            "compression_ratio": 1.7545787545787546,
            "no_speech_prob": 0.0031223297119140625,
            "words": [
                {
                    "start": 555.05,
                    "end": 555.11,
                    "word": " we",
                    "probability": 0.1673583984375
                },
                {
                    "start": 555.11,
                    "end": 555.25,
                    "word": " need",
                    "probability": 0.7685546875
                },
                {
                    "start": 555.25,
                    "end": 555.73,
                    "word": " algorithms",
                    "probability": 0.8818359375
                },
                {
                    "start": 555.73,
                    "end": 556.13,
                    "word": " that",
                    "probability": 0.87353515625
                },
                {
                    "start": 556.13,
                    "end": 556.29,
                    "word": " can",
                    "probability": 0.89111328125
                },
                {
                    "start": 556.29,
                    "end": 556.47,
                    "word": " make",
                    "probability": 0.87744140625
                },
                {
                    "start": 556.47,
                    "end": 556.73,
                    "word": " sense",
                    "probability": 0.74560546875
                },
                {
                    "start": 556.73,
                    "end": 556.87,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 556.87,
                    "end": 557.27,
                    "word": " language",
                    "probability": 0.68115234375
                },
                {
                    "start": 557.27,
                    "end": 557.55,
                    "word": ".",
                    "probability": 0.67578125
                },
                {
                    "start": 557.55,
                    "end": 557.77,
                    "word": " Making",
                    "probability": 0.78857421875
                },
                {
                    "start": 557.77,
                    "end": 557.95,
                    "word": " sense",
                    "probability": 0.75244140625
                },
                {
                    "start": 557.95,
                    "end": 558.05,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 558.05,
                    "end": 558.25,
                    "word": " language",
                    "probability": 0.68017578125
                },
                {
                    "start": 558.25,
                    "end": 558.41,
                    "word": " is",
                    "probability": 0.89697265625
                },
                {
                    "start": 558.41,
                    "end": 558.69,
                    "word": " hard",
                    "probability": 0.79150390625
                },
                {
                    "start": 558.69,
                    "end": 558.77,
                    "word": ",",
                    "probability": 0.432373046875
                },
                {
                    "start": 558.77,
                    "end": 558.93,
                    "word": " as",
                    "probability": 0.85595703125
                },
                {
                    "start": 558.93,
                    "end": 559.07,
                    "word": " these",
                    "probability": 0.71484375
                },
                {
                    "start": 559.07,
                    "end": 559.33,
                    "word": " three",
                    "probability": 0.5849609375
                }
            ]
        },
        {
            "id": 92,
            "seek": 56922,
            "start": 559.33,
            "end": 568.61,
            "text": " problems demonstrate. So a good solution to the phishing problem would model as intermediate",
            "tokens": [
                50580,
                2740,
                11698,
                13,
                407,
                257,
                665,
                3827,
                281,
                264,
                903,
                3807,
                1154,
                576,
                2316,
                382,
                19376,
                51048
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26487379521131516,
            "compression_ratio": 1.7545787545787546,
            "no_speech_prob": 0.0031223297119140625,
            "words": [
                {
                    "start": 559.33,
                    "end": 559.71,
                    "word": " problems",
                    "probability": 0.68896484375
                },
                {
                    "start": 559.71,
                    "end": 560.23,
                    "word": " demonstrate",
                    "probability": 0.81298828125
                },
                {
                    "start": 560.23,
                    "end": 560.67,
                    "word": ".",
                    "probability": 0.912109375
                },
                {
                    "start": 562.19,
                    "end": 562.63,
                    "word": " So",
                    "probability": 0.8134765625
                },
                {
                    "start": 562.75,
                    "end": 562.87,
                    "word": " a good",
                    "probability": 0.88330078125
                },
                {
                    "start": 562.87,
                    "end": 563.49,
                    "word": " solution",
                    "probability": 0.76806640625
                },
                {
                    "start": 563.49,
                    "end": 563.65,
                    "word": " to",
                    "probability": 0.8935546875
                },
                {
                    "start": 563.65,
                    "end": 563.75,
                    "word": " the",
                    "probability": 0.81689453125
                },
                {
                    "start": 563.75,
                    "end": 563.93,
                    "word": " phishing",
                    "probability": 0.87109375
                },
                {
                    "start": 563.93,
                    "end": 564.37,
                    "word": " problem",
                    "probability": 0.7763671875
                },
                {
                    "start": 564.37,
                    "end": 565.27,
                    "word": " would",
                    "probability": 0.818359375
                },
                {
                    "start": 565.27,
                    "end": 565.89,
                    "word": " model",
                    "probability": 0.80615234375
                },
                {
                    "start": 565.89,
                    "end": 568.21,
                    "word": " as",
                    "probability": 0.79345703125
                },
                {
                    "start": 568.21,
                    "end": 568.61,
                    "word": " intermediate",
                    "probability": 0.7646484375
                }
            ]
        },
        {
            "id": 93,
            "seek": 56922,
            "start": 568.61,
            "end": 574.15,
            "text": " steps to detecting that an email is a phishing email. At some level, be able to solve these",
            "tokens": [
                51048,
                4439,
                281,
                40237,
                300,
                364,
                3796,
                307,
                257,
                903,
                3807,
                3796,
                13,
                1711,
                512,
                1496,
                11,
                312,
                1075,
                281,
                5039,
                613,
                51320
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26487379521131516,
            "compression_ratio": 1.7545787545787546,
            "no_speech_prob": 0.0031223297119140625,
            "words": [
                {
                    "start": 568.61,
                    "end": 569.03,
                    "word": " steps",
                    "probability": 0.72216796875
                },
                {
                    "start": 569.03,
                    "end": 569.15,
                    "word": " to",
                    "probability": 0.38525390625
                },
                {
                    "start": 569.15,
                    "end": 569.47,
                    "word": " detecting",
                    "probability": 0.7529296875
                },
                {
                    "start": 569.47,
                    "end": 570.01,
                    "word": " that",
                    "probability": 0.8740234375
                },
                {
                    "start": 570.01,
                    "end": 570.87,
                    "word": " an",
                    "probability": 0.59228515625
                },
                {
                    "start": 570.87,
                    "end": 571.49,
                    "word": " email",
                    "probability": 0.86181640625
                },
                {
                    "start": 571.49,
                    "end": 571.65,
                    "word": " is",
                    "probability": 0.84716796875
                },
                {
                    "start": 571.75,
                    "end": 571.95,
                    "word": " a phishing",
                    "probability": 0.911865234375
                },
                {
                    "start": 571.95,
                    "end": 572.27,
                    "word": " email",
                    "probability": 0.92138671875
                },
                {
                    "start": 572.27,
                    "end": 572.57,
                    "word": ".",
                    "probability": 0.60595703125
                },
                {
                    "start": 572.57,
                    "end": 572.75,
                    "word": " At",
                    "probability": 0.8798828125
                },
                {
                    "start": 572.75,
                    "end": 572.91,
                    "word": " some",
                    "probability": 0.81201171875
                },
                {
                    "start": 572.91,
                    "end": 573.19,
                    "word": " level",
                    "probability": 0.82177734375
                },
                {
                    "start": 573.19,
                    "end": 573.31,
                    "word": ",",
                    "probability": 0.83447265625
                },
                {
                    "start": 573.31,
                    "end": 573.53,
                    "word": " be",
                    "probability": 0.7841796875
                },
                {
                    "start": 573.53,
                    "end": 573.67,
                    "word": " able",
                    "probability": 0.8974609375
                },
                {
                    "start": 573.67,
                    "end": 573.81,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 573.81,
                    "end": 573.97,
                    "word": " solve",
                    "probability": 0.833984375
                },
                {
                    "start": 573.97,
                    "end": 574.15,
                    "word": " these",
                    "probability": 0.69677734375
                }
            ]
        },
        {
            "id": 94,
            "seek": 56922,
            "start": 574.15,
            "end": 577.93,
            "text": " problems and sit somewhere in the depths of this intermediate representation of the language that",
            "tokens": [
                51320,
                2740,
                293,
                1394,
                4079,
                294,
                264,
                28439,
                295,
                341,
                19376,
                10290,
                295,
                264,
                2856,
                300,
                51508
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26487379521131516,
            "compression_ratio": 1.7545787545787546,
            "no_speech_prob": 0.0031223297119140625,
            "words": [
                {
                    "start": 574.15,
                    "end": 574.51,
                    "word": " problems",
                    "probability": 0.68212890625
                },
                {
                    "start": 574.51,
                    "end": 574.81,
                    "word": " and",
                    "probability": 0.26123046875
                },
                {
                    "start": 574.81,
                    "end": 575.03,
                    "word": " sit",
                    "probability": 0.2529296875
                },
                {
                    "start": 575.03,
                    "end": 575.27,
                    "word": " somewhere",
                    "probability": 0.77490234375
                },
                {
                    "start": 575.27,
                    "end": 575.53,
                    "word": " in",
                    "probability": 0.8974609375
                },
                {
                    "start": 575.53,
                    "end": 575.73,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 575.73,
                    "end": 575.95,
                    "word": " depths",
                    "probability": 0.6787109375
                },
                {
                    "start": 575.95,
                    "end": 576.09,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 576.09,
                    "end": 576.23,
                    "word": " this",
                    "probability": 0.607421875
                },
                {
                    "start": 576.23,
                    "end": 576.61,
                    "word": " intermediate",
                    "probability": 0.72509765625
                },
                {
                    "start": 576.61,
                    "end": 577.09,
                    "word": " representation",
                    "probability": 0.74609375
                },
                {
                    "start": 577.09,
                    "end": 577.43,
                    "word": " of",
                    "probability": 0.90478515625
                },
                {
                    "start": 577.43,
                    "end": 577.55,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 577.55,
                    "end": 577.75,
                    "word": " language",
                    "probability": 0.67041015625
                },
                {
                    "start": 577.75,
                    "end": 577.93,
                    "word": " that",
                    "probability": 0.8583984375
                }
            ]
        },
        {
            "id": 95,
            "seek": 56922,
            "start": 577.93,
            "end": 582.41,
            "text": " it's looking at. The other challenge we have, obviously, in any cybersecurity problem, at least",
            "tokens": [
                51508,
                309,
                311,
                1237,
                412,
                13,
                440,
                661,
                3430,
                321,
                362,
                11,
                2745,
                11,
                294,
                604,
                38765,
                1154,
                11,
                412,
                1935,
                51732
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26487379521131516,
            "compression_ratio": 1.7545787545787546,
            "no_speech_prob": 0.0031223297119140625,
            "words": [
                {
                    "start": 577.93,
                    "end": 578.05,
                    "word": " it's",
                    "probability": 0.90771484375
                },
                {
                    "start": 578.05,
                    "end": 578.29,
                    "word": " looking",
                    "probability": 0.79541015625
                },
                {
                    "start": 578.29,
                    "end": 578.51,
                    "word": " at",
                    "probability": 0.9033203125
                },
                {
                    "start": 578.51,
                    "end": 578.95,
                    "word": ".",
                    "probability": 0.99755859375
                },
                {
                    "start": 579.17,
                    "end": 579.49,
                    "word": " The",
                    "probability": 0.76611328125
                },
                {
                    "start": 579.49,
                    "end": 579.61,
                    "word": " other",
                    "probability": 0.8037109375
                },
                {
                    "start": 579.61,
                    "end": 579.85,
                    "word": " challenge",
                    "probability": 0.759765625
                },
                {
                    "start": 579.85,
                    "end": 579.97,
                    "word": " we",
                    "probability": 0.46533203125
                },
                {
                    "start": 579.97,
                    "end": 580.17,
                    "word": " have",
                    "probability": 0.8720703125
                },
                {
                    "start": 580.17,
                    "end": 580.25,
                    "word": ",",
                    "probability": 0.6162109375
                },
                {
                    "start": 580.25,
                    "end": 580.49,
                    "word": " obviously",
                    "probability": 0.77099609375
                },
                {
                    "start": 580.49,
                    "end": 580.59,
                    "word": ",",
                    "probability": 0.8740234375
                },
                {
                    "start": 580.59,
                    "end": 580.71,
                    "word": " in",
                    "probability": 0.89306640625
                },
                {
                    "start": 580.71,
                    "end": 581.05,
                    "word": " any",
                    "probability": 0.841796875
                },
                {
                    "start": 581.05,
                    "end": 581.41,
                    "word": " cybersecurity",
                    "probability": 0.63818359375
                },
                {
                    "start": 581.41,
                    "end": 581.93,
                    "word": " problem",
                    "probability": 0.76220703125
                },
                {
                    "start": 581.93,
                    "end": 582.11,
                    "word": ",",
                    "probability": 0.75244140625
                },
                {
                    "start": 582.11,
                    "end": 582.27,
                    "word": " at",
                    "probability": 0.410888671875
                },
                {
                    "start": 582.27,
                    "end": 582.41,
                    "word": " least",
                    "probability": 0.841796875
                }
            ]
        },
        {
            "id": 96,
            "seek": 59684,
            "start": 582.41,
            "end": 586.35,
            "text": " any detection context in cybersecurity, is that we have adversaries who'd like to bypass our",
            "tokens": [
                50364,
                604,
                17784,
                4319,
                294,
                38765,
                11,
                307,
                300,
                321,
                362,
                17641,
                4889,
                567,
                1116,
                411,
                281,
                24996,
                527,
                50560
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2319711590042481,
            "compression_ratio": 1.6859205776173285,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 582.41,
                    "end": 582.55,
                    "word": " any",
                    "probability": 0.275634765625
                },
                {
                    "start": 582.55,
                    "end": 582.95,
                    "word": " detection",
                    "probability": 0.6904296875
                },
                {
                    "start": 582.95,
                    "end": 583.69,
                    "word": " context",
                    "probability": 0.7158203125
                },
                {
                    "start": 583.69,
                    "end": 583.85,
                    "word": " in",
                    "probability": 0.796875
                },
                {
                    "start": 583.85,
                    "end": 584.23,
                    "word": " cybersecurity",
                    "probability": 0.62353515625
                },
                {
                    "start": 584.23,
                    "end": 584.45,
                    "word": ",",
                    "probability": 0.09832763671875
                },
                {
                    "start": 584.45,
                    "end": 584.49,
                    "word": " is",
                    "probability": 0.82666015625
                },
                {
                    "start": 584.49,
                    "end": 584.67,
                    "word": " that",
                    "probability": 0.8662109375
                },
                {
                    "start": 584.67,
                    "end": 584.91,
                    "word": " we",
                    "probability": 0.89404296875
                },
                {
                    "start": 584.91,
                    "end": 585.03,
                    "word": " have",
                    "probability": 0.87744140625
                },
                {
                    "start": 585.03,
                    "end": 585.37,
                    "word": " adversaries",
                    "probability": 0.909423828125
                },
                {
                    "start": 585.37,
                    "end": 585.61,
                    "word": " who'd",
                    "probability": 0.73974609375
                },
                {
                    "start": 585.61,
                    "end": 585.73,
                    "word": " like",
                    "probability": 0.88037109375
                },
                {
                    "start": 585.73,
                    "end": 585.87,
                    "word": " to",
                    "probability": 0.900390625
                },
                {
                    "start": 585.87,
                    "end": 586.07,
                    "word": " bypass",
                    "probability": 0.8427734375
                },
                {
                    "start": 586.07,
                    "end": 586.35,
                    "word": " our",
                    "probability": 0.80517578125
                }
            ]
        },
        {
            "id": 97,
            "seek": 59684,
            "start": 586.35,
            "end": 591.49,
            "text": " detection. And that's also worth considering. So these are all reasons why the problem that we're",
            "tokens": [
                50560,
                17784,
                13,
                400,
                300,
                311,
                611,
                3163,
                8079,
                13,
                407,
                613,
                366,
                439,
                4112,
                983,
                264,
                1154,
                300,
                321,
                434,
                50824
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2319711590042481,
            "compression_ratio": 1.6859205776173285,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 586.35,
                    "end": 586.69,
                    "word": " detection",
                    "probability": 0.7734375
                },
                {
                    "start": 586.69,
                    "end": 587.09,
                    "word": ".",
                    "probability": 0.29931640625
                },
                {
                    "start": 587.09,
                    "end": 587.13,
                    "word": " And",
                    "probability": 0.8662109375
                },
                {
                    "start": 587.13,
                    "end": 587.29,
                    "word": " that's",
                    "probability": 0.931884765625
                },
                {
                    "start": 587.29,
                    "end": 587.47,
                    "word": " also",
                    "probability": 0.830078125
                },
                {
                    "start": 587.47,
                    "end": 587.67,
                    "word": " worth",
                    "probability": 0.90673828125
                },
                {
                    "start": 587.67,
                    "end": 588.05,
                    "word": " considering",
                    "probability": 0.7421875
                },
                {
                    "start": 588.05,
                    "end": 588.43,
                    "word": ".",
                    "probability": 0.98291015625
                },
                {
                    "start": 588.43,
                    "end": 588.51,
                    "word": " So",
                    "probability": 0.6259765625
                },
                {
                    "start": 588.51,
                    "end": 588.63,
                    "word": " these",
                    "probability": 0.6181640625
                },
                {
                    "start": 588.63,
                    "end": 588.73,
                    "word": " are",
                    "probability": 0.892578125
                },
                {
                    "start": 588.73,
                    "end": 588.95,
                    "word": " all",
                    "probability": 0.8984375
                },
                {
                    "start": 588.95,
                    "end": 589.31,
                    "word": " reasons",
                    "probability": 0.8701171875
                },
                {
                    "start": 589.31,
                    "end": 589.59,
                    "word": " why",
                    "probability": 0.80517578125
                },
                {
                    "start": 589.59,
                    "end": 590.69,
                    "word": " the",
                    "probability": 0.759765625
                },
                {
                    "start": 590.69,
                    "end": 591.01,
                    "word": " problem",
                    "probability": 0.7373046875
                },
                {
                    "start": 591.01,
                    "end": 591.17,
                    "word": " that",
                    "probability": 0.87646484375
                },
                {
                    "start": 591.17,
                    "end": 591.49,
                    "word": " we're",
                    "probability": 0.93310546875
                }
            ]
        },
        {
            "id": 98,
            "seek": 59684,
            "start": 591.49,
            "end": 596.19,
            "text": " presenting here is a hard one, deserves... we haven't solved it completely, deserves attention",
            "tokens": [
                50824,
                15578,
                510,
                307,
                257,
                1152,
                472,
                11,
                17037,
                485,
                321,
                2378,
                380,
                13041,
                309,
                2584,
                11,
                17037,
                3202,
                51052
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2319711590042481,
            "compression_ratio": 1.6859205776173285,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 591.49,
                    "end": 591.97,
                    "word": " presenting",
                    "probability": 0.705078125
                },
                {
                    "start": 591.97,
                    "end": 592.25,
                    "word": " here",
                    "probability": 0.8154296875
                },
                {
                    "start": 592.25,
                    "end": 592.41,
                    "word": " is",
                    "probability": 0.892578125
                },
                {
                    "start": 592.61,
                    "end": 592.77,
                    "word": " a hard",
                    "probability": 0.830078125
                },
                {
                    "start": 592.77,
                    "end": 593.03,
                    "word": " one",
                    "probability": 0.81787109375
                },
                {
                    "start": 593.03,
                    "end": 593.41,
                    "word": ",",
                    "probability": 0.50830078125
                },
                {
                    "start": 593.41,
                    "end": 594.03,
                    "word": " deserves...",
                    "probability": 0.3123779296875
                },
                {
                    "start": 594.03,
                    "end": 594.29,
                    "word": " we",
                    "probability": 0.2376708984375
                },
                {
                    "start": 594.29,
                    "end": 594.45,
                    "word": " haven't",
                    "probability": 0.929931640625
                },
                {
                    "start": 594.45,
                    "end": 594.63,
                    "word": " solved",
                    "probability": 0.7626953125
                },
                {
                    "start": 594.63,
                    "end": 594.75,
                    "word": " it",
                    "probability": 0.8974609375
                },
                {
                    "start": 594.75,
                    "end": 595.09,
                    "word": " completely",
                    "probability": 0.6513671875
                },
                {
                    "start": 595.09,
                    "end": 595.43,
                    "word": ",",
                    "probability": 0.802734375
                },
                {
                    "start": 595.43,
                    "end": 595.83,
                    "word": " deserves",
                    "probability": 0.75341796875
                },
                {
                    "start": 595.83,
                    "end": 596.19,
                    "word": " attention",
                    "probability": 0.8720703125
                }
            ]
        },
        {
            "id": 99,
            "seek": 59684,
            "start": 596.19,
            "end": 605.11,
            "text": " from our community. Now, the approach that we're using to attack the phishing problem",
            "tokens": [
                51052,
                490,
                527,
                1768,
                13,
                823,
                11,
                264,
                3109,
                300,
                321,
                434,
                1228,
                281,
                2690,
                264,
                903,
                3807,
                1154,
                51452
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2319711590042481,
            "compression_ratio": 1.6859205776173285,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 596.19,
                    "end": 596.41,
                    "word": " from",
                    "probability": 0.8173828125
                },
                {
                    "start": 596.41,
                    "end": 596.59,
                    "word": " our",
                    "probability": 0.80078125
                },
                {
                    "start": 596.59,
                    "end": 597.23,
                    "word": " community",
                    "probability": 0.771484375
                },
                {
                    "start": 597.23,
                    "end": 597.75,
                    "word": ".",
                    "probability": 0.99658203125
                },
                {
                    "start": 600.45,
                    "end": 600.97,
                    "word": " Now",
                    "probability": 0.8740234375
                },
                {
                    "start": 600.97,
                    "end": 601.45,
                    "word": ",",
                    "probability": 0.83349609375
                },
                {
                    "start": 601.45,
                    "end": 601.83,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 601.83,
                    "end": 602.19,
                    "word": " approach",
                    "probability": 0.79345703125
                },
                {
                    "start": 602.19,
                    "end": 602.35,
                    "word": " that",
                    "probability": 0.8818359375
                },
                {
                    "start": 602.35,
                    "end": 602.49,
                    "word": " we're",
                    "probability": 0.9326171875
                },
                {
                    "start": 602.49,
                    "end": 602.83,
                    "word": " using",
                    "probability": 0.83154296875
                },
                {
                    "start": 602.83,
                    "end": 603.91,
                    "word": " to",
                    "probability": 0.86083984375
                },
                {
                    "start": 603.91,
                    "end": 604.27,
                    "word": " attack",
                    "probability": 0.830078125
                },
                {
                    "start": 604.27,
                    "end": 604.45,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 604.45,
                    "end": 604.69,
                    "word": " phishing",
                    "probability": 0.869873046875
                },
                {
                    "start": 604.69,
                    "end": 605.11,
                    "word": " problem",
                    "probability": 0.77392578125
                }
            ]
        },
        {
            "id": 100,
            "seek": 59684,
            "start": 605.61,
            "end": 611.49,
            "text": " is based in neural networks and deep learning, and a specific advance that happened in the last",
            "tokens": [
                51452,
                307,
                2361,
                294,
                18161,
                9590,
                293,
                2452,
                2539,
                11,
                293,
                257,
                2685,
                7295,
                300,
                2011,
                294,
                264,
                1036,
                51744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2319711590042481,
            "compression_ratio": 1.6859205776173285,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 605.61,
                    "end": 605.87,
                    "word": " is",
                    "probability": 0.86181640625
                },
                {
                    "start": 605.87,
                    "end": 606.11,
                    "word": " based",
                    "probability": 0.81005859375
                },
                {
                    "start": 606.11,
                    "end": 606.33,
                    "word": " in",
                    "probability": 0.7392578125
                },
                {
                    "start": 606.33,
                    "end": 606.59,
                    "word": " neural",
                    "probability": 0.84521484375
                },
                {
                    "start": 606.59,
                    "end": 606.89,
                    "word": " networks",
                    "probability": 0.66796875
                },
                {
                    "start": 606.89,
                    "end": 607.05,
                    "word": " and",
                    "probability": 0.85205078125
                },
                {
                    "start": 607.05,
                    "end": 607.23,
                    "word": " deep",
                    "probability": 0.91357421875
                },
                {
                    "start": 607.23,
                    "end": 607.53,
                    "word": " learning",
                    "probability": 0.82421875
                },
                {
                    "start": 607.53,
                    "end": 607.79,
                    "word": ",",
                    "probability": 0.35595703125
                },
                {
                    "start": 607.79,
                    "end": 608.31,
                    "word": " and",
                    "probability": 0.689453125
                },
                {
                    "start": 608.57,
                    "end": 608.97,
                    "word": " a specific",
                    "probability": 0.8017578125
                },
                {
                    "start": 608.97,
                    "end": 609.79,
                    "word": " advance",
                    "probability": 0.87451171875
                },
                {
                    "start": 609.79,
                    "end": 610.17,
                    "word": " that",
                    "probability": 0.875
                },
                {
                    "start": 610.17,
                    "end": 610.55,
                    "word": " happened",
                    "probability": 0.54248046875
                },
                {
                    "start": 610.55,
                    "end": 611.19,
                    "word": " in",
                    "probability": 0.88330078125
                },
                {
                    "start": 611.19,
                    "end": 611.31,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 611.31,
                    "end": 611.49,
                    "word": " last",
                    "probability": 0.81201171875
                }
            ]
        },
        {
            "id": 101,
            "seek": 62178,
            "start": 611.49,
            "end": 619.71,
            "text": " few years known as transformers. So transformers, or more specifically, transformer blocks, are a",
            "tokens": [
                50364,
                1326,
                924,
                2570,
                382,
                4088,
                433,
                13,
                407,
                4088,
                433,
                11,
                420,
                544,
                4682,
                11,
                31782,
                8474,
                11,
                366,
                257,
                50760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21661184116413718,
            "compression_ratio": 1.69377990430622,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 611.49,
                    "end": 611.67,
                    "word": " few",
                    "probability": 0.359619140625
                },
                {
                    "start": 611.67,
                    "end": 612.05,
                    "word": " years",
                    "probability": 0.880859375
                },
                {
                    "start": 612.05,
                    "end": 613.09,
                    "word": " known",
                    "probability": 0.61669921875
                },
                {
                    "start": 613.09,
                    "end": 613.67,
                    "word": " as",
                    "probability": 0.9111328125
                },
                {
                    "start": 613.67,
                    "end": 614.59,
                    "word": " transformers",
                    "probability": 0.795166015625
                },
                {
                    "start": 614.59,
                    "end": 615.07,
                    "word": ".",
                    "probability": 0.93310546875
                },
                {
                    "start": 615.51,
                    "end": 615.99,
                    "word": " So",
                    "probability": 0.75
                },
                {
                    "start": 615.99,
                    "end": 616.79,
                    "word": " transformers",
                    "probability": 0.809326171875
                },
                {
                    "start": 616.79,
                    "end": 617.05,
                    "word": ",",
                    "probability": 0.4599609375
                },
                {
                    "start": 617.05,
                    "end": 617.27,
                    "word": " or",
                    "probability": 0.8701171875
                },
                {
                    "start": 617.27,
                    "end": 617.71,
                    "word": " more",
                    "probability": 0.84619140625
                },
                {
                    "start": 617.71,
                    "end": 618.05,
                    "word": " specifically",
                    "probability": 0.72412109375
                },
                {
                    "start": 618.05,
                    "end": 618.25,
                    "word": ",",
                    "probability": 0.2391357421875
                },
                {
                    "start": 618.25,
                    "end": 618.53,
                    "word": " transformer",
                    "probability": 0.8359375
                },
                {
                    "start": 618.53,
                    "end": 618.99,
                    "word": " blocks",
                    "probability": 0.80615234375
                },
                {
                    "start": 618.99,
                    "end": 619.19,
                    "word": ",",
                    "probability": 0.58349609375
                },
                {
                    "start": 619.19,
                    "end": 619.37,
                    "word": " are",
                    "probability": 0.85009765625
                },
                {
                    "start": 619.57,
                    "end": 619.71,
                    "word": " a new",
                    "probability": 0.80859375
                }
            ]
        },
        {
            "id": 102,
            "seek": 62178,
            "start": 619.71,
            "end": 626.95,
            "text": " new kind of construct in neural networks, much like convolutions were new, I think, as of the",
            "tokens": [
                50760,
                777,
                733,
                295,
                7690,
                294,
                18161,
                9590,
                11,
                709,
                411,
                3754,
                15892,
                645,
                777,
                11,
                286,
                519,
                11,
                382,
                295,
                264,
                51112
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21661184116413718,
            "compression_ratio": 1.69377990430622,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 619.71,
                    "end": 619.87,
                    "word": " kind",
                    "probability": 0.78466796875
                },
                {
                    "start": 619.87,
                    "end": 619.99,
                    "word": " of",
                    "probability": 0.91162109375
                },
                {
                    "start": 619.99,
                    "end": 620.45,
                    "word": " construct",
                    "probability": 0.9208984375
                },
                {
                    "start": 620.45,
                    "end": 621.05,
                    "word": " in",
                    "probability": 0.888671875
                },
                {
                    "start": 621.05,
                    "end": 622.15,
                    "word": " neural",
                    "probability": 0.8564453125
                },
                {
                    "start": 622.15,
                    "end": 622.55,
                    "word": " networks",
                    "probability": 0.65869140625
                },
                {
                    "start": 622.55,
                    "end": 622.77,
                    "word": ",",
                    "probability": 0.6689453125
                },
                {
                    "start": 622.77,
                    "end": 623.07,
                    "word": " much",
                    "probability": 0.80322265625
                },
                {
                    "start": 623.07,
                    "end": 623.37,
                    "word": " like",
                    "probability": 0.89111328125
                },
                {
                    "start": 623.37,
                    "end": 624.83,
                    "word": " convolutions",
                    "probability": 0.7275390625
                },
                {
                    "start": 624.83,
                    "end": 625.11,
                    "word": " were",
                    "probability": 0.85107421875
                },
                {
                    "start": 625.11,
                    "end": 625.29,
                    "word": " new",
                    "probability": 0.810546875
                },
                {
                    "start": 625.29,
                    "end": 625.37,
                    "word": ",",
                    "probability": 0.66943359375
                },
                {
                    "start": 625.37,
                    "end": 625.43,
                    "word": " I",
                    "probability": 0.99267578125
                },
                {
                    "start": 625.43,
                    "end": 625.55,
                    "word": " think",
                    "probability": 0.8046875
                },
                {
                    "start": 625.55,
                    "end": 625.57,
                    "word": ",",
                    "probability": 0.8427734375
                },
                {
                    "start": 625.57,
                    "end": 625.71,
                    "word": " as",
                    "probability": 0.90185546875
                },
                {
                    "start": 625.71,
                    "end": 625.87,
                    "word": " of",
                    "probability": 0.8916015625
                },
                {
                    "start": 625.87,
                    "end": 626.13,
                    "word": " the",
                    "probability": 0.8046875
                },
                {
                    "start": 626.13,
                    "end": 626.95,
                    "word": " 90s",
                    "probability": 0.7142333984375
                }
            ]
        },
        {
            "id": 103,
            "seek": 62178,
            "start": 626.95,
            "end": 631.73,
            "text": " 90s or late 80s, and backpropagation was a new idea in neural networks, I think,",
            "tokens": [
                51112,
                4289,
                82,
                420,
                3469,
                4688,
                82,
                11,
                293,
                646,
                79,
                1513,
                559,
                399,
                390,
                257,
                777,
                1558,
                294,
                18161,
                9590,
                11,
                286,
                519,
                11,
                51352
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21661184116413718,
            "compression_ratio": 1.69377990430622,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 626.95,
                    "end": 627.07,
                    "word": " or",
                    "probability": 0.8212890625
                },
                {
                    "start": 627.07,
                    "end": 627.51,
                    "word": " late",
                    "probability": 0.84619140625
                },
                {
                    "start": 627.51,
                    "end": 627.89,
                    "word": " 80s",
                    "probability": 0.95068359375
                },
                {
                    "start": 627.89,
                    "end": 627.97,
                    "word": ",",
                    "probability": 0.537109375
                },
                {
                    "start": 627.97,
                    "end": 628.09,
                    "word": " and",
                    "probability": 0.88525390625
                },
                {
                    "start": 628.09,
                    "end": 629.49,
                    "word": " backpropagation",
                    "probability": 0.86611328125
                },
                {
                    "start": 629.49,
                    "end": 629.67,
                    "word": " was",
                    "probability": 0.87841796875
                },
                {
                    "start": 629.79,
                    "end": 630.17,
                    "word": " a new",
                    "probability": 0.7685546875
                },
                {
                    "start": 630.17,
                    "end": 630.41,
                    "word": " idea",
                    "probability": 0.77001953125
                },
                {
                    "start": 630.41,
                    "end": 630.53,
                    "word": " in",
                    "probability": 0.85400390625
                },
                {
                    "start": 630.53,
                    "end": 630.69,
                    "word": " neural",
                    "probability": 0.884765625
                },
                {
                    "start": 630.69,
                    "end": 631.01,
                    "word": " networks",
                    "probability": 0.6435546875
                },
                {
                    "start": 631.01,
                    "end": 631.07,
                    "word": ",",
                    "probability": 0.80517578125
                },
                {
                    "start": 631.07,
                    "end": 631.15,
                    "word": " I",
                    "probability": 0.95556640625
                },
                {
                    "start": 631.15,
                    "end": 631.25,
                    "word": " think",
                    "probability": 0.80322265625
                },
                {
                    "start": 631.25,
                    "end": 631.33,
                    "word": ",",
                    "probability": 0.861328125
                },
                {
                    "start": 631.33,
                    "end": 631.53,
                    "word": " starting",
                    "probability": 0.81689453125
                },
                {
                    "start": 631.53,
                    "end": 631.73,
                    "word": " in",
                    "probability": 0.900390625
                }
            ]
        },
        {
            "id": 104,
            "seek": 62178,
            "start": 631.73,
            "end": 636.43,
            "text": " starting in the 80s. Transformers are a new idea that's come out in recent years,",
            "tokens": [
                51352,
                2891,
                294,
                264,
                4688,
                82,
                13,
                27938,
                433,
                366,
                257,
                777,
                1558,
                300,
                311,
                808,
                484,
                294,
                5162,
                924,
                11,
                51580
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21661184116413718,
            "compression_ratio": 1.69377990430622,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 631.73,
                    "end": 632.13,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 632.13,
                    "end": 632.55,
                    "word": " 80s",
                    "probability": 0.939697265625
                },
                {
                    "start": 632.55,
                    "end": 632.87,
                    "word": ".",
                    "probability": 0.94677734375
                },
                {
                    "start": 632.87,
                    "end": 633.33,
                    "word": " Transformers",
                    "probability": 0.921630859375
                },
                {
                    "start": 633.33,
                    "end": 633.47,
                    "word": " are",
                    "probability": 0.8828125
                },
                {
                    "start": 633.67,
                    "end": 633.77,
                    "word": " a new",
                    "probability": 0.81396484375
                },
                {
                    "start": 633.77,
                    "end": 634.01,
                    "word": " idea",
                    "probability": 0.77587890625
                },
                {
                    "start": 634.01,
                    "end": 634.33,
                    "word": " that's",
                    "probability": 0.931396484375
                },
                {
                    "start": 634.33,
                    "end": 634.59,
                    "word": " come",
                    "probability": 0.81884765625
                },
                {
                    "start": 634.59,
                    "end": 634.81,
                    "word": " out",
                    "probability": 0.82470703125
                },
                {
                    "start": 634.81,
                    "end": 635.03,
                    "word": " in",
                    "probability": 0.90771484375
                },
                {
                    "start": 635.03,
                    "end": 635.47,
                    "word": " recent",
                    "probability": 0.79248046875
                },
                {
                    "start": 635.47,
                    "end": 635.89,
                    "word": " years",
                    "probability": 0.8857421875
                },
                {
                    "start": 635.89,
                    "end": 636.43,
                    "word": ",",
                    "probability": 0.462158203125
                }
            ]
        },
        {
            "id": 105,
            "seek": 64762,
            "start": 637.29,
            "end": 644.37,
            "text": " and they help model language with a depth and fidelity that seems to be genuinely new and",
            "tokens": [
                50412,
                293,
                436,
                854,
                2316,
                2856,
                365,
                257,
                7161,
                293,
                46404,
                300,
                2544,
                281,
                312,
                17839,
                777,
                293,
                50756
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18068181344053963,
            "compression_ratio": 1.7168458781362008,
            "no_speech_prob": 0.00040435791015625,
            "words": [
                {
                    "start": 637.29,
                    "end": 637.73,
                    "word": " and",
                    "probability": 0.30908203125
                },
                {
                    "start": 637.73,
                    "end": 638.17,
                    "word": " they",
                    "probability": 0.734375
                },
                {
                    "start": 638.17,
                    "end": 638.39,
                    "word": " help",
                    "probability": 0.78125
                },
                {
                    "start": 638.39,
                    "end": 638.65,
                    "word": " model",
                    "probability": 0.8154296875
                },
                {
                    "start": 638.65,
                    "end": 639.09,
                    "word": " language",
                    "probability": 0.6806640625
                },
                {
                    "start": 639.09,
                    "end": 639.97,
                    "word": " with",
                    "probability": 0.796875
                },
                {
                    "start": 640.21,
                    "end": 640.53,
                    "word": " a depth",
                    "probability": 0.82568359375
                },
                {
                    "start": 640.53,
                    "end": 640.91,
                    "word": " and",
                    "probability": 0.8876953125
                },
                {
                    "start": 640.91,
                    "end": 641.33,
                    "word": " fidelity",
                    "probability": 0.77490234375
                },
                {
                    "start": 641.33,
                    "end": 642.29,
                    "word": " that",
                    "probability": 0.85595703125
                },
                {
                    "start": 642.29,
                    "end": 642.97,
                    "word": " seems",
                    "probability": 0.76171875
                },
                {
                    "start": 642.97,
                    "end": 643.15,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 643.15,
                    "end": 643.37,
                    "word": " be",
                    "probability": 0.90625
                },
                {
                    "start": 643.37,
                    "end": 643.87,
                    "word": " genuinely",
                    "probability": 0.72607421875
                },
                {
                    "start": 643.87,
                    "end": 644.23,
                    "word": " new",
                    "probability": 0.7998046875
                },
                {
                    "start": 644.23,
                    "end": 644.37,
                    "word": " and",
                    "probability": 0.8486328125
                }
            ]
        },
        {
            "id": 106,
            "seek": 64762,
            "start": 644.37,
            "end": 649.09,
            "text": " represent a step function in our ability to model language. So they're very exciting. The big idea",
            "tokens": [
                50756,
                2906,
                257,
                1823,
                2445,
                294,
                527,
                3485,
                281,
                2316,
                2856,
                13,
                407,
                436,
                434,
                588,
                4670,
                13,
                440,
                955,
                1558,
                50996
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18068181344053963,
            "compression_ratio": 1.7168458781362008,
            "no_speech_prob": 0.00040435791015625,
            "words": [
                {
                    "start": 644.37,
                    "end": 644.63,
                    "word": " represent",
                    "probability": 0.7109375
                },
                {
                    "start": 644.81,
                    "end": 644.99,
                    "word": " a step",
                    "probability": 0.80078125
                },
                {
                    "start": 644.99,
                    "end": 645.39,
                    "word": " function",
                    "probability": 0.8515625
                },
                {
                    "start": 645.39,
                    "end": 645.75,
                    "word": " in",
                    "probability": 0.85107421875
                },
                {
                    "start": 645.75,
                    "end": 645.85,
                    "word": " our",
                    "probability": 0.810546875
                },
                {
                    "start": 645.85,
                    "end": 646.11,
                    "word": " ability",
                    "probability": 0.86376953125
                },
                {
                    "start": 646.11,
                    "end": 646.29,
                    "word": " to",
                    "probability": 0.90771484375
                },
                {
                    "start": 646.29,
                    "end": 646.47,
                    "word": " model",
                    "probability": 0.82861328125
                },
                {
                    "start": 646.47,
                    "end": 646.91,
                    "word": " language",
                    "probability": 0.6708984375
                },
                {
                    "start": 646.91,
                    "end": 647.23,
                    "word": ".",
                    "probability": 0.7734375
                },
                {
                    "start": 647.23,
                    "end": 647.41,
                    "word": " So",
                    "probability": 0.830078125
                },
                {
                    "start": 647.41,
                    "end": 647.53,
                    "word": " they're",
                    "probability": 0.78466796875
                },
                {
                    "start": 647.53,
                    "end": 647.71,
                    "word": " very",
                    "probability": 0.7412109375
                },
                {
                    "start": 647.71,
                    "end": 648.03,
                    "word": " exciting",
                    "probability": 0.861328125
                },
                {
                    "start": 648.03,
                    "end": 648.39,
                    "word": ".",
                    "probability": 0.88330078125
                },
                {
                    "start": 648.39,
                    "end": 648.67,
                    "word": " The",
                    "probability": 0.80078125
                },
                {
                    "start": 648.67,
                    "end": 648.83,
                    "word": " big",
                    "probability": 0.88134765625
                },
                {
                    "start": 648.83,
                    "end": 649.09,
                    "word": " idea",
                    "probability": 0.77880859375
                }
            ]
        },
        {
            "id": 107,
            "seek": 64762,
            "start": 649.09,
            "end": 652.67,
            "text": " behind the work that we're presenting today is that we're taking transformers and applying them",
            "tokens": [
                50996,
                2261,
                264,
                589,
                300,
                321,
                434,
                15578,
                965,
                307,
                300,
                321,
                434,
                1940,
                4088,
                433,
                293,
                9275,
                552,
                51172
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18068181344053963,
            "compression_ratio": 1.7168458781362008,
            "no_speech_prob": 0.00040435791015625,
            "words": [
                {
                    "start": 649.09,
                    "end": 649.41,
                    "word": " behind",
                    "probability": 0.849609375
                },
                {
                    "start": 649.41,
                    "end": 649.61,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 649.61,
                    "end": 649.75,
                    "word": " work",
                    "probability": 0.79736328125
                },
                {
                    "start": 649.75,
                    "end": 649.83,
                    "word": " that",
                    "probability": 0.83544921875
                },
                {
                    "start": 649.83,
                    "end": 649.93,
                    "word": " we're",
                    "probability": 0.917724609375
                },
                {
                    "start": 649.93,
                    "end": 650.15,
                    "word": " presenting",
                    "probability": 0.40625
                },
                {
                    "start": 650.15,
                    "end": 650.41,
                    "word": " today",
                    "probability": 0.798828125
                },
                {
                    "start": 650.41,
                    "end": 650.53,
                    "word": " is",
                    "probability": 0.88427734375
                },
                {
                    "start": 650.53,
                    "end": 650.65,
                    "word": " that",
                    "probability": 0.87841796875
                },
                {
                    "start": 650.65,
                    "end": 650.75,
                    "word": " we're",
                    "probability": 0.9296875
                },
                {
                    "start": 650.75,
                    "end": 651.03,
                    "word": " taking",
                    "probability": 0.810546875
                },
                {
                    "start": 651.03,
                    "end": 651.71,
                    "word": " transformers",
                    "probability": 0.7978515625
                },
                {
                    "start": 651.71,
                    "end": 652.27,
                    "word": " and",
                    "probability": 0.87939453125
                },
                {
                    "start": 652.27,
                    "end": 652.51,
                    "word": " applying",
                    "probability": 0.70703125
                },
                {
                    "start": 652.51,
                    "end": 652.67,
                    "word": " them",
                    "probability": 0.7158203125
                }
            ]
        },
        {
            "id": 108,
            "seek": 64762,
            "start": 652.67,
            "end": 657.01,
            "text": " to a cybersecurity problem, which we haven't seen much of before. So I'm going to talk a bit about",
            "tokens": [
                51172,
                281,
                257,
                38765,
                1154,
                11,
                597,
                321,
                2378,
                380,
                1612,
                709,
                295,
                949,
                13,
                407,
                286,
                478,
                516,
                281,
                751,
                257,
                857,
                466,
                51388
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18068181344053963,
            "compression_ratio": 1.7168458781362008,
            "no_speech_prob": 0.00040435791015625,
            "words": [
                {
                    "start": 652.67,
                    "end": 652.77,
                    "word": " to",
                    "probability": 0.89208984375
                },
                {
                    "start": 652.91,
                    "end": 653.23,
                    "word": " a cybersecurity",
                    "probability": 0.66455078125
                },
                {
                    "start": 653.23,
                    "end": 653.71,
                    "word": " problem",
                    "probability": 0.76171875
                },
                {
                    "start": 653.71,
                    "end": 653.83,
                    "word": ",",
                    "probability": 0.53662109375
                },
                {
                    "start": 653.83,
                    "end": 653.89,
                    "word": " which",
                    "probability": 0.91357421875
                },
                {
                    "start": 653.89,
                    "end": 654.01,
                    "word": " we",
                    "probability": 0.90673828125
                },
                {
                    "start": 654.01,
                    "end": 654.19,
                    "word": " haven't",
                    "probability": 0.93603515625
                },
                {
                    "start": 654.19,
                    "end": 654.37,
                    "word": " seen",
                    "probability": 0.87548828125
                },
                {
                    "start": 654.37,
                    "end": 654.57,
                    "word": " much",
                    "probability": 0.80908203125
                },
                {
                    "start": 654.57,
                    "end": 654.75,
                    "word": " of",
                    "probability": 0.8984375
                },
                {
                    "start": 654.75,
                    "end": 655.31,
                    "word": " before",
                    "probability": 0.7841796875
                },
                {
                    "start": 655.31,
                    "end": 655.75,
                    "word": ".",
                    "probability": 0.9921875
                },
                {
                    "start": 655.81,
                    "end": 656.23,
                    "word": " So",
                    "probability": 0.8759765625
                },
                {
                    "start": 656.23,
                    "end": 656.41,
                    "word": " I'm",
                    "probability": 0.973876953125
                },
                {
                    "start": 656.41,
                    "end": 656.43,
                    "word": " going",
                    "probability": 0.443359375
                },
                {
                    "start": 656.43,
                    "end": 656.49,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 656.49,
                    "end": 656.65,
                    "word": " talk",
                    "probability": 0.80712890625
                },
                {
                    "start": 656.75,
                    "end": 656.83,
                    "word": " a bit",
                    "probability": 0.8974609375
                },
                {
                    "start": 656.83,
                    "end": 657.01,
                    "word": " about",
                    "probability": 0.81982421875
                }
            ]
        },
        {
            "id": 109,
            "seek": 64762,
            "start": 657.01,
            "end": 662.27,
            "text": " what transformers are. A detailed discussion of how they work is beyond the scope of this talk,",
            "tokens": [
                51388,
                437,
                4088,
                433,
                366,
                13,
                316,
                9942,
                5017,
                295,
                577,
                436,
                589,
                307,
                4399,
                264,
                11923,
                295,
                341,
                751,
                11,
                51648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18068181344053963,
            "compression_ratio": 1.7168458781362008,
            "no_speech_prob": 0.00040435791015625,
            "words": [
                {
                    "start": 657.01,
                    "end": 657.17,
                    "word": " what",
                    "probability": 0.87451171875
                },
                {
                    "start": 657.17,
                    "end": 657.69,
                    "word": " transformers",
                    "probability": 0.92919921875
                },
                {
                    "start": 657.69,
                    "end": 657.97,
                    "word": " are",
                    "probability": 0.8955078125
                },
                {
                    "start": 657.97,
                    "end": 658.41,
                    "word": ".",
                    "probability": 0.72314453125
                },
                {
                    "start": 659.07,
                    "end": 659.51,
                    "word": " A",
                    "probability": 0.93212890625
                },
                {
                    "start": 659.51,
                    "end": 659.81,
                    "word": " detailed",
                    "probability": 0.68798828125
                },
                {
                    "start": 659.81,
                    "end": 660.19,
                    "word": " discussion",
                    "probability": 0.75244140625
                },
                {
                    "start": 660.19,
                    "end": 660.51,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 660.51,
                    "end": 660.79,
                    "word": " how",
                    "probability": 0.873046875
                },
                {
                    "start": 660.79,
                    "end": 660.93,
                    "word": " they",
                    "probability": 0.73779296875
                },
                {
                    "start": 660.93,
                    "end": 661.11,
                    "word": " work",
                    "probability": 0.8154296875
                },
                {
                    "start": 661.11,
                    "end": 661.25,
                    "word": " is",
                    "probability": 0.89599609375
                },
                {
                    "start": 661.25,
                    "end": 661.43,
                    "word": " beyond",
                    "probability": 0.7373046875
                },
                {
                    "start": 661.43,
                    "end": 661.57,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 661.57,
                    "end": 661.71,
                    "word": " scope",
                    "probability": 0.8310546875
                },
                {
                    "start": 661.71,
                    "end": 661.83,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 661.83,
                    "end": 661.97,
                    "word": " this",
                    "probability": 0.880859375
                },
                {
                    "start": 661.97,
                    "end": 662.21,
                    "word": " talk",
                    "probability": 0.8125
                },
                {
                    "start": 662.21,
                    "end": 662.27,
                    "word": ",",
                    "probability": 0.8623046875
                }
            ]
        },
        {
            "id": 110,
            "seek": 67760,
            "start": 662.27,
            "end": 669.13,
            "text": " but I'm going to give some intuition. And then I'll pass the mic metaphorically over to",
            "tokens": [
                50364,
                457,
                286,
                478,
                516,
                281,
                976,
                512,
                24002,
                13,
                400,
                550,
                286,
                603,
                1320,
                264,
                3123,
                19157,
                984,
                670,
                281,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23275861383854657,
            "compression_ratio": 1.5775862068965518,
            "no_speech_prob": 0.0016689300537109375,
            "words": [
                {
                    "start": 662.27,
                    "end": 662.37,
                    "word": " but",
                    "probability": 0.350341796875
                },
                {
                    "start": 662.37,
                    "end": 662.49,
                    "word": " I'm",
                    "probability": 0.979736328125
                },
                {
                    "start": 662.49,
                    "end": 662.53,
                    "word": " going",
                    "probability": 0.463134765625
                },
                {
                    "start": 662.53,
                    "end": 662.57,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 662.57,
                    "end": 662.69,
                    "word": " give",
                    "probability": 0.7763671875
                },
                {
                    "start": 662.69,
                    "end": 662.81,
                    "word": " some",
                    "probability": 0.779296875
                },
                {
                    "start": 662.81,
                    "end": 663.17,
                    "word": " intuition",
                    "probability": 0.80859375
                },
                {
                    "start": 663.17,
                    "end": 663.73,
                    "word": ".",
                    "probability": 0.291015625
                },
                {
                    "start": 664.15,
                    "end": 664.15,
                    "word": " And",
                    "probability": 0.7509765625
                },
                {
                    "start": 664.15,
                    "end": 664.37,
                    "word": " then",
                    "probability": 0.72705078125
                },
                {
                    "start": 664.37,
                    "end": 665.03,
                    "word": " I'll",
                    "probability": 0.834228515625
                },
                {
                    "start": 665.03,
                    "end": 665.51,
                    "word": " pass",
                    "probability": 0.8251953125
                },
                {
                    "start": 665.51,
                    "end": 666.85,
                    "word": " the",
                    "probability": 0.76708984375
                },
                {
                    "start": 666.85,
                    "end": 667.09,
                    "word": " mic",
                    "probability": 0.83642578125
                },
                {
                    "start": 667.09,
                    "end": 668.31,
                    "word": " metaphorically",
                    "probability": 0.769287109375
                },
                {
                    "start": 668.31,
                    "end": 668.67,
                    "word": " over",
                    "probability": 0.84912109375
                },
                {
                    "start": 668.67,
                    "end": 669.13,
                    "word": " to",
                    "probability": 0.9130859375
                }
            ]
        },
        {
            "id": 111,
            "seek": 67760,
            "start": 669.13,
            "end": 679.54,
            "text": " Young, who will present on how we're using transformers. Okay, so here's an example,",
            "tokens": [
                50704,
                8160,
                11,
                567,
                486,
                1974,
                322,
                577,
                321,
                434,
                1228,
                4088,
                433,
                13,
                1033,
                11,
                370,
                510,
                311,
                364,
                1365,
                11,
                51100
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23275861383854657,
            "compression_ratio": 1.5775862068965518,
            "no_speech_prob": 0.0016689300537109375,
            "words": [
                {
                    "start": 669.13,
                    "end": 669.41,
                    "word": " Young",
                    "probability": 0.42236328125
                },
                {
                    "start": 669.41,
                    "end": 669.45,
                    "word": ",",
                    "probability": 0.297607421875
                },
                {
                    "start": 669.45,
                    "end": 669.61,
                    "word": " who",
                    "probability": 0.8134765625
                },
                {
                    "start": 669.61,
                    "end": 670.03,
                    "word": " will",
                    "probability": 0.78955078125
                },
                {
                    "start": 670.03,
                    "end": 670.83,
                    "word": " present",
                    "probability": 0.72412109375
                },
                {
                    "start": 670.83,
                    "end": 671.33,
                    "word": " on",
                    "probability": 0.888671875
                },
                {
                    "start": 671.33,
                    "end": 671.93,
                    "word": " how",
                    "probability": 0.890625
                },
                {
                    "start": 671.93,
                    "end": 672.07,
                    "word": " we're",
                    "probability": 0.9189453125
                },
                {
                    "start": 672.07,
                    "end": 672.29,
                    "word": " using",
                    "probability": 0.83349609375
                },
                {
                    "start": 672.29,
                    "end": 672.99,
                    "word": " transformers",
                    "probability": 0.796875
                },
                {
                    "start": 672.99,
                    "end": 673.55,
                    "word": ".",
                    "probability": 0.9931640625
                },
                {
                    "start": 676.9,
                    "end": 677.46,
                    "word": " Okay",
                    "probability": 0.3662109375
                },
                {
                    "start": 677.46,
                    "end": 677.92,
                    "word": ",",
                    "probability": 0.5712890625
                },
                {
                    "start": 677.92,
                    "end": 678.16,
                    "word": " so",
                    "probability": 0.88232421875
                },
                {
                    "start": 678.16,
                    "end": 678.86,
                    "word": " here's",
                    "probability": 0.87841796875
                },
                {
                    "start": 678.86,
                    "end": 678.96,
                    "word": " an",
                    "probability": 0.90380859375
                },
                {
                    "start": 678.96,
                    "end": 679.26,
                    "word": " example",
                    "probability": 0.92431640625
                },
                {
                    "start": 679.26,
                    "end": 679.54,
                    "word": ",",
                    "probability": 0.428466796875
                }
            ]
        },
        {
            "id": 112,
            "seek": 67760,
            "start": 679.54,
            "end": 686.94,
            "text": " which I think helps to illustrate some ways in which transformers represent a real breakthrough",
            "tokens": [
                51100,
                597,
                286,
                519,
                3665,
                281,
                23221,
                512,
                2098,
                294,
                597,
                4088,
                433,
                2906,
                257,
                957,
                22397,
                51500
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23275861383854657,
            "compression_ratio": 1.5775862068965518,
            "no_speech_prob": 0.0016689300537109375,
            "words": [
                {
                    "start": 679.54,
                    "end": 679.62,
                    "word": " which",
                    "probability": 0.91357421875
                },
                {
                    "start": 679.62,
                    "end": 679.74,
                    "word": " I",
                    "probability": 0.994140625
                },
                {
                    "start": 679.74,
                    "end": 679.9,
                    "word": " think",
                    "probability": 0.81494140625
                },
                {
                    "start": 679.9,
                    "end": 680.34,
                    "word": " helps",
                    "probability": 0.73876953125
                },
                {
                    "start": 680.34,
                    "end": 680.56,
                    "word": " to",
                    "probability": 0.89892578125
                },
                {
                    "start": 680.56,
                    "end": 680.96,
                    "word": " illustrate",
                    "probability": 0.86669921875
                },
                {
                    "start": 680.96,
                    "end": 683.2,
                    "word": " some",
                    "probability": 0.77392578125
                },
                {
                    "start": 683.2,
                    "end": 683.44,
                    "word": " ways",
                    "probability": 0.81689453125
                },
                {
                    "start": 683.44,
                    "end": 683.58,
                    "word": " in",
                    "probability": 0.9033203125
                },
                {
                    "start": 683.58,
                    "end": 683.76,
                    "word": " which",
                    "probability": 0.90185546875
                },
                {
                    "start": 683.76,
                    "end": 684.56,
                    "word": " transformers",
                    "probability": 0.9140625
                },
                {
                    "start": 684.56,
                    "end": 685.66,
                    "word": " represent",
                    "probability": 0.73046875
                },
                {
                    "start": 686.08,
                    "end": 686.54,
                    "word": " a real",
                    "probability": 0.87548828125
                },
                {
                    "start": 686.54,
                    "end": 686.94,
                    "word": " breakthrough",
                    "probability": 0.8173828125
                }
            ]
        },
        {
            "id": 113,
            "seek": 67760,
            "start": 686.94,
            "end": 694.58,
            "text": " in machine comprehension of language. So in this example, and I expect some of you have seen this",
            "tokens": [
                51500,
                294,
                3479,
                44991,
                295,
                2856,
                13,
                407,
                294,
                341,
                1365,
                11,
                293,
                286,
                2066,
                512,
                295,
                291,
                362,
                1612,
                341,
                51860
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23275861383854657,
            "compression_ratio": 1.5775862068965518,
            "no_speech_prob": 0.0016689300537109375,
            "words": [
                {
                    "start": 686.94,
                    "end": 687.62,
                    "word": " in",
                    "probability": 0.58154296875
                },
                {
                    "start": 687.62,
                    "end": 688.62,
                    "word": " machine",
                    "probability": 0.69970703125
                },
                {
                    "start": 688.62,
                    "end": 689.94,
                    "word": " comprehension",
                    "probability": 0.8271484375
                },
                {
                    "start": 689.94,
                    "end": 690.28,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 690.28,
                    "end": 690.92,
                    "word": " language",
                    "probability": 0.64208984375
                },
                {
                    "start": 690.92,
                    "end": 691.48,
                    "word": ".",
                    "probability": 0.99560546875
                },
                {
                    "start": 691.78,
                    "end": 692.0,
                    "word": " So",
                    "probability": 0.8125
                },
                {
                    "start": 692.0,
                    "end": 692.34,
                    "word": " in",
                    "probability": 0.77783203125
                },
                {
                    "start": 692.34,
                    "end": 692.52,
                    "word": " this",
                    "probability": 0.8740234375
                },
                {
                    "start": 692.52,
                    "end": 692.86,
                    "word": " example",
                    "probability": 0.92041015625
                },
                {
                    "start": 692.86,
                    "end": 693.12,
                    "word": ",",
                    "probability": 0.892578125
                },
                {
                    "start": 693.12,
                    "end": 693.3,
                    "word": " and",
                    "probability": 0.81982421875
                },
                {
                    "start": 693.3,
                    "end": 693.5,
                    "word": " I",
                    "probability": 0.9716796875
                },
                {
                    "start": 693.5,
                    "end": 693.74,
                    "word": " expect",
                    "probability": 0.81884765625
                },
                {
                    "start": 693.74,
                    "end": 693.92,
                    "word": " some",
                    "probability": 0.8046875
                },
                {
                    "start": 693.92,
                    "end": 694.02,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 694.02,
                    "end": 694.14,
                    "word": " you",
                    "probability": 0.89892578125
                },
                {
                    "start": 694.14,
                    "end": 694.28,
                    "word": " have",
                    "probability": 0.87451171875
                },
                {
                    "start": 694.28,
                    "end": 694.46,
                    "word": " seen",
                    "probability": 0.876953125
                },
                {
                    "start": 694.46,
                    "end": 694.58,
                    "word": " this",
                    "probability": 0.87890625
                }
            ]
        },
        {
            "id": 114,
            "seek": 70176,
            "start": 694.58,
            "end": 702.72,
            "text": " because this went viral last year, a researcher wrote a prompt. This prompt is given at the top",
            "tokens": [
                50364,
                570,
                341,
                1437,
                16132,
                1036,
                1064,
                11,
                257,
                21751,
                4114,
                257,
                12391,
                13,
                639,
                12391,
                307,
                2212,
                412,
                264,
                1192,
                50768
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2116168495537578,
            "compression_ratio": 1.5815217391304348,
            "no_speech_prob": 0.0003142356872558594,
            "words": [
                {
                    "start": 694.58,
                    "end": 694.76,
                    "word": " because",
                    "probability": 0.243896484375
                },
                {
                    "start": 694.76,
                    "end": 695.04,
                    "word": " this",
                    "probability": 0.87353515625
                },
                {
                    "start": 695.04,
                    "end": 695.48,
                    "word": " went",
                    "probability": 0.1436767578125
                },
                {
                    "start": 695.48,
                    "end": 696.02,
                    "word": " viral",
                    "probability": 0.81201171875
                },
                {
                    "start": 696.02,
                    "end": 697.04,
                    "word": " last",
                    "probability": 0.787109375
                },
                {
                    "start": 697.04,
                    "end": 697.32,
                    "word": " year",
                    "probability": 0.8857421875
                },
                {
                    "start": 697.32,
                    "end": 698.02,
                    "word": ",",
                    "probability": 0.3193359375
                },
                {
                    "start": 698.7,
                    "end": 699.0,
                    "word": " a researcher",
                    "probability": 0.80419921875
                },
                {
                    "start": 699.0,
                    "end": 699.44,
                    "word": " wrote",
                    "probability": 0.85302734375
                },
                {
                    "start": 699.64,
                    "end": 699.98,
                    "word": " a prompt",
                    "probability": 0.84619140625
                },
                {
                    "start": 699.98,
                    "end": 700.54,
                    "word": ".",
                    "probability": 0.72900390625
                },
                {
                    "start": 700.54,
                    "end": 701.14,
                    "word": " This",
                    "probability": 0.82275390625
                },
                {
                    "start": 701.14,
                    "end": 701.38,
                    "word": " prompt",
                    "probability": 0.8271484375
                },
                {
                    "start": 701.38,
                    "end": 701.52,
                    "word": " is",
                    "probability": 0.8896484375
                },
                {
                    "start": 701.52,
                    "end": 701.78,
                    "word": " given",
                    "probability": 0.79833984375
                },
                {
                    "start": 701.78,
                    "end": 702.34,
                    "word": " at",
                    "probability": 0.908203125
                },
                {
                    "start": 702.34,
                    "end": 702.5,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 702.5,
                    "end": 702.72,
                    "word": " top",
                    "probability": 0.81640625
                }
            ]
        },
        {
            "id": 115,
            "seek": 70176,
            "start": 702.72,
            "end": 710.36,
            "text": " here in a shocking finding, that paragraph. And then a transformer model just took it from there",
            "tokens": [
                50768,
                510,
                294,
                257,
                18776,
                5006,
                11,
                300,
                18865,
                13,
                400,
                550,
                257,
                31782,
                2316,
                445,
                1890,
                309,
                490,
                456,
                51172
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2116168495537578,
            "compression_ratio": 1.5815217391304348,
            "no_speech_prob": 0.0003142356872558594,
            "words": [
                {
                    "start": 702.72,
                    "end": 703.02,
                    "word": " here",
                    "probability": 0.81103515625
                },
                {
                    "start": 703.02,
                    "end": 704.08,
                    "word": " in",
                    "probability": 0.65478515625
                },
                {
                    "start": 704.18,
                    "end": 704.4,
                    "word": " a shocking",
                    "probability": 0.7236328125
                },
                {
                    "start": 704.4,
                    "end": 704.82,
                    "word": " finding",
                    "probability": 0.7158203125
                },
                {
                    "start": 704.82,
                    "end": 705.0,
                    "word": ",",
                    "probability": 0.63525390625
                },
                {
                    "start": 705.0,
                    "end": 705.08,
                    "word": " that",
                    "probability": 0.88134765625
                },
                {
                    "start": 705.08,
                    "end": 705.58,
                    "word": " paragraph",
                    "probability": 0.79736328125
                },
                {
                    "start": 705.58,
                    "end": 706.18,
                    "word": ".",
                    "probability": 0.70556640625
                },
                {
                    "start": 706.72,
                    "end": 706.82,
                    "word": " And",
                    "probability": 0.8017578125
                },
                {
                    "start": 706.82,
                    "end": 707.1,
                    "word": " then",
                    "probability": 0.7177734375
                },
                {
                    "start": 707.42,
                    "end": 707.76,
                    "word": " a transformer",
                    "probability": 0.75341796875
                },
                {
                    "start": 707.76,
                    "end": 708.2,
                    "word": " model",
                    "probability": 0.796875
                },
                {
                    "start": 708.2,
                    "end": 709.4,
                    "word": " just",
                    "probability": 0.8076171875
                },
                {
                    "start": 709.4,
                    "end": 709.8,
                    "word": " took",
                    "probability": 0.73681640625
                },
                {
                    "start": 709.8,
                    "end": 709.94,
                    "word": " it",
                    "probability": 0.90087890625
                },
                {
                    "start": 709.94,
                    "end": 710.1,
                    "word": " from",
                    "probability": 0.81494140625
                },
                {
                    "start": 710.1,
                    "end": 710.36,
                    "word": " there",
                    "probability": 0.853515625
                }
            ]
        },
        {
            "id": 116,
            "seek": 70176,
            "start": 710.36,
            "end": 718.74,
            "text": " and wrote a story based on that prompt. And I think when you see the story that the machine wrote,",
            "tokens": [
                51172,
                293,
                4114,
                257,
                1657,
                2361,
                322,
                300,
                12391,
                13,
                400,
                286,
                519,
                562,
                291,
                536,
                264,
                1657,
                300,
                264,
                3479,
                4114,
                11,
                51584
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2116168495537578,
            "compression_ratio": 1.5815217391304348,
            "no_speech_prob": 0.0003142356872558594,
            "words": [
                {
                    "start": 710.36,
                    "end": 711.2,
                    "word": " and",
                    "probability": 0.84521484375
                },
                {
                    "start": 711.2,
                    "end": 712.5,
                    "word": " wrote",
                    "probability": 0.8681640625
                },
                {
                    "start": 713.04,
                    "end": 713.46,
                    "word": " a story",
                    "probability": 0.884765625
                },
                {
                    "start": 713.46,
                    "end": 713.82,
                    "word": " based",
                    "probability": 0.8193359375
                },
                {
                    "start": 713.82,
                    "end": 713.94,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 713.94,
                    "end": 714.1,
                    "word": " that",
                    "probability": 0.88134765625
                },
                {
                    "start": 714.1,
                    "end": 714.4,
                    "word": " prompt",
                    "probability": 0.8427734375
                },
                {
                    "start": 714.4,
                    "end": 715.0,
                    "word": ".",
                    "probability": 0.99072265625
                },
                {
                    "start": 715.78,
                    "end": 715.9,
                    "word": " And",
                    "probability": 0.759765625
                },
                {
                    "start": 715.9,
                    "end": 716.0,
                    "word": " I",
                    "probability": 0.9892578125
                },
                {
                    "start": 716.0,
                    "end": 716.14,
                    "word": " think",
                    "probability": 0.80908203125
                },
                {
                    "start": 716.14,
                    "end": 716.56,
                    "word": " when",
                    "probability": 0.84521484375
                },
                {
                    "start": 716.56,
                    "end": 716.74,
                    "word": " you",
                    "probability": 0.89599609375
                },
                {
                    "start": 716.74,
                    "end": 717.16,
                    "word": " see",
                    "probability": 0.80126953125
                },
                {
                    "start": 717.16,
                    "end": 717.34,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 717.34,
                    "end": 717.58,
                    "word": " story",
                    "probability": 0.87744140625
                },
                {
                    "start": 717.58,
                    "end": 717.78,
                    "word": " that",
                    "probability": 0.87548828125
                },
                {
                    "start": 717.78,
                    "end": 717.94,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 717.94,
                    "end": 718.18,
                    "word": " machine",
                    "probability": 0.72900390625
                },
                {
                    "start": 718.18,
                    "end": 718.54,
                    "word": " wrote",
                    "probability": 0.853515625
                },
                {
                    "start": 718.54,
                    "end": 718.74,
                    "word": ",",
                    "probability": 0.916015625
                }
            ]
        },
        {
            "id": 117,
            "seek": 73086,
            "start": 718.74,
            "end": 726.68,
            "text": " you'll see that it betrays a understanding of the syntax and semantics of language",
            "tokens": [
                50372,
                291,
                603,
                536,
                300,
                309,
                15560,
                82,
                257,
                3701,
                295,
                264,
                28431,
                293,
                4361,
                45298,
                295,
                2856,
                50760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20943509285839704,
            "compression_ratio": 1.6115107913669064,
            "no_speech_prob": 0.0006556510925292969,
            "words": [
                {
                    "start": 718.74,
                    "end": 719.22,
                    "word": " you'll",
                    "probability": 0.7012939453125
                },
                {
                    "start": 719.22,
                    "end": 719.38,
                    "word": " see",
                    "probability": 0.8203125
                },
                {
                    "start": 719.38,
                    "end": 719.6,
                    "word": " that",
                    "probability": 0.87353515625
                },
                {
                    "start": 719.6,
                    "end": 719.92,
                    "word": " it",
                    "probability": 0.76416015625
                },
                {
                    "start": 719.92,
                    "end": 720.64,
                    "word": " betrays",
                    "probability": 0.80810546875
                },
                {
                    "start": 723.28,
                    "end": 724.66,
                    "word": " a understanding",
                    "probability": 0.68994140625
                },
                {
                    "start": 724.66,
                    "end": 724.82,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 724.82,
                    "end": 724.94,
                    "word": " the",
                    "probability": 0.82275390625
                },
                {
                    "start": 724.94,
                    "end": 725.24,
                    "word": " syntax",
                    "probability": 0.9189453125
                },
                {
                    "start": 725.24,
                    "end": 725.46,
                    "word": " and",
                    "probability": 0.89404296875
                },
                {
                    "start": 725.46,
                    "end": 725.8,
                    "word": " semantics",
                    "probability": 0.95068359375
                },
                {
                    "start": 725.8,
                    "end": 726.08,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 726.08,
                    "end": 726.68,
                    "word": " language",
                    "probability": 0.64404296875
                }
            ]
        },
        {
            "id": 118,
            "seek": 73086,
            "start": 727.76,
            "end": 732.56,
            "text": " that's quite striking. So I think it's worth reading this out loud. The human-written prompt",
            "tokens": [
                50816,
                300,
                311,
                1596,
                18559,
                13,
                407,
                286,
                519,
                309,
                311,
                3163,
                3760,
                341,
                484,
                6588,
                13,
                440,
                1952,
                12,
                26859,
                12391,
                51060
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20943509285839704,
            "compression_ratio": 1.6115107913669064,
            "no_speech_prob": 0.0006556510925292969,
            "words": [
                {
                    "start": 727.76,
                    "end": 728.22,
                    "word": " that's",
                    "probability": 0.848388671875
                },
                {
                    "start": 728.22,
                    "end": 728.42,
                    "word": " quite",
                    "probability": 0.88818359375
                },
                {
                    "start": 728.42,
                    "end": 728.82,
                    "word": " striking",
                    "probability": 0.81201171875
                },
                {
                    "start": 728.82,
                    "end": 729.28,
                    "word": ".",
                    "probability": 0.9423828125
                },
                {
                    "start": 729.58,
                    "end": 729.72,
                    "word": " So",
                    "probability": 0.8232421875
                },
                {
                    "start": 729.72,
                    "end": 729.8,
                    "word": " I",
                    "probability": 0.83740234375
                },
                {
                    "start": 729.8,
                    "end": 729.86,
                    "word": " think",
                    "probability": 0.81298828125
                },
                {
                    "start": 729.86,
                    "end": 729.98,
                    "word": " it's",
                    "probability": 0.949951171875
                },
                {
                    "start": 729.98,
                    "end": 730.1,
                    "word": " worth",
                    "probability": 0.87841796875
                },
                {
                    "start": 730.1,
                    "end": 730.32,
                    "word": " reading",
                    "probability": 0.80126953125
                },
                {
                    "start": 730.32,
                    "end": 730.48,
                    "word": " this",
                    "probability": 0.88037109375
                },
                {
                    "start": 730.48,
                    "end": 730.64,
                    "word": " out",
                    "probability": 0.8076171875
                },
                {
                    "start": 730.64,
                    "end": 730.8,
                    "word": " loud",
                    "probability": 0.90771484375
                },
                {
                    "start": 730.8,
                    "end": 731.26,
                    "word": ".",
                    "probability": 0.84912109375
                },
                {
                    "start": 731.36,
                    "end": 731.6,
                    "word": " The",
                    "probability": 0.77392578125
                },
                {
                    "start": 731.6,
                    "end": 731.96,
                    "word": " human",
                    "probability": 0.66796875
                },
                {
                    "start": 731.96,
                    "end": 732.14,
                    "word": "-written",
                    "probability": 0.425537109375
                },
                {
                    "start": 732.14,
                    "end": 732.56,
                    "word": " prompt",
                    "probability": 0.7978515625
                }
            ]
        },
        {
            "id": 119,
            "seek": 73086,
            "start": 732.56,
            "end": 737.86,
            "text": " reads, in a shocking finding, scientists discovered a herd of unicorns living in a remote,",
            "tokens": [
                51060,
                15700,
                11,
                294,
                257,
                18776,
                5006,
                11,
                7708,
                6941,
                257,
                29484,
                295,
                28122,
                82,
                2647,
                294,
                257,
                8607,
                11,
                51308
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20943509285839704,
            "compression_ratio": 1.6115107913669064,
            "no_speech_prob": 0.0006556510925292969,
            "words": [
                {
                    "start": 732.56,
                    "end": 733.06,
                    "word": " reads",
                    "probability": 0.755859375
                },
                {
                    "start": 733.06,
                    "end": 733.22,
                    "word": ",",
                    "probability": 0.79150390625
                },
                {
                    "start": 733.22,
                    "end": 733.26,
                    "word": " in",
                    "probability": 0.10980224609375
                },
                {
                    "start": 733.36,
                    "end": 733.64,
                    "word": " a shocking",
                    "probability": 0.6611328125
                },
                {
                    "start": 733.64,
                    "end": 734.16,
                    "word": " finding",
                    "probability": 0.72998046875
                },
                {
                    "start": 734.16,
                    "end": 735.04,
                    "word": ",",
                    "probability": 0.8095703125
                },
                {
                    "start": 735.04,
                    "end": 735.5,
                    "word": " scientists",
                    "probability": 0.5732421875
                },
                {
                    "start": 735.5,
                    "end": 735.94,
                    "word": " discovered",
                    "probability": 0.6201171875
                },
                {
                    "start": 736.16,
                    "end": 736.3,
                    "word": " a herd",
                    "probability": 0.798828125
                },
                {
                    "start": 736.3,
                    "end": 736.46,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 736.46,
                    "end": 736.98,
                    "word": " unicorns",
                    "probability": 0.831298828125
                },
                {
                    "start": 736.98,
                    "end": 737.14,
                    "word": " living",
                    "probability": 0.72802734375
                },
                {
                    "start": 737.14,
                    "end": 737.28,
                    "word": " in",
                    "probability": 0.90283203125
                },
                {
                    "start": 737.36,
                    "end": 737.66,
                    "word": " a remote",
                    "probability": 0.87646484375
                },
                {
                    "start": 737.66,
                    "end": 737.86,
                    "word": ",",
                    "probability": 0.68896484375
                }
            ]
        },
        {
            "id": 120,
            "seek": 73086,
            "start": 737.86,
            "end": 742.36,
            "text": " previously unexplored valley in the Andes Mountains. Even more surprising to the researchers",
            "tokens": [
                51308,
                8046,
                11572,
                564,
                2769,
                17636,
                294,
                264,
                400,
                279,
                30970,
                13,
                2754,
                544,
                8830,
                281,
                264,
                10309,
                51544
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20943509285839704,
            "compression_ratio": 1.6115107913669064,
            "no_speech_prob": 0.0006556510925292969,
            "words": [
                {
                    "start": 737.86,
                    "end": 738.1,
                    "word": " previously",
                    "probability": 0.7294921875
                },
                {
                    "start": 738.1,
                    "end": 738.62,
                    "word": " unexplored",
                    "probability": 0.9064127604166666
                },
                {
                    "start": 738.62,
                    "end": 738.96,
                    "word": " valley",
                    "probability": 0.69384765625
                },
                {
                    "start": 738.96,
                    "end": 739.14,
                    "word": " in",
                    "probability": 0.89013671875
                },
                {
                    "start": 739.14,
                    "end": 739.24,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 739.24,
                    "end": 739.52,
                    "word": " Andes",
                    "probability": 0.9228515625
                },
                {
                    "start": 739.52,
                    "end": 740.0,
                    "word": " Mountains",
                    "probability": 0.546875
                },
                {
                    "start": 740.0,
                    "end": 740.46,
                    "word": ".",
                    "probability": 0.97314453125
                },
                {
                    "start": 740.78,
                    "end": 741.08,
                    "word": " Even",
                    "probability": 0.80908203125
                },
                {
                    "start": 741.08,
                    "end": 741.3,
                    "word": " more",
                    "probability": 0.8759765625
                },
                {
                    "start": 741.3,
                    "end": 741.68,
                    "word": " surprising",
                    "probability": 0.81640625
                },
                {
                    "start": 741.68,
                    "end": 741.84,
                    "word": " to",
                    "probability": 0.90869140625
                },
                {
                    "start": 741.84,
                    "end": 742.0,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 742.0,
                    "end": 742.36,
                    "word": " researchers",
                    "probability": 0.7568359375
                }
            ]
        },
        {
            "id": 121,
            "seek": 73086,
            "start": 742.36,
            "end": 747.84,
            "text": " was the fact that the unicorn spoke perfect English. And now again, the challenge to the",
            "tokens": [
                51544,
                390,
                264,
                1186,
                300,
                264,
                28122,
                7179,
                2176,
                3669,
                13,
                400,
                586,
                797,
                11,
                264,
                3430,
                281,
                264,
                51740
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20943509285839704,
            "compression_ratio": 1.6115107913669064,
            "no_speech_prob": 0.0006556510925292969,
            "words": [
                {
                    "start": 742.36,
                    "end": 742.64,
                    "word": " was",
                    "probability": 0.880859375
                },
                {
                    "start": 742.64,
                    "end": 742.78,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 742.78,
                    "end": 742.98,
                    "word": " fact",
                    "probability": 0.80126953125
                },
                {
                    "start": 742.98,
                    "end": 743.1,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 743.1,
                    "end": 743.24,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 743.24,
                    "end": 743.6,
                    "word": " unicorn",
                    "probability": 0.75732421875
                },
                {
                    "start": 743.6,
                    "end": 744.02,
                    "word": " spoke",
                    "probability": 0.326171875
                },
                {
                    "start": 744.02,
                    "end": 744.48,
                    "word": " perfect",
                    "probability": 0.865234375
                },
                {
                    "start": 744.48,
                    "end": 745.0,
                    "word": " English",
                    "probability": 0.90966796875
                },
                {
                    "start": 745.0,
                    "end": 745.34,
                    "word": ".",
                    "probability": 0.93310546875
                },
                {
                    "start": 745.34,
                    "end": 745.48,
                    "word": " And",
                    "probability": 0.837890625
                },
                {
                    "start": 745.48,
                    "end": 745.6,
                    "word": " now",
                    "probability": 0.7919921875
                },
                {
                    "start": 745.6,
                    "end": 745.76,
                    "word": " again",
                    "probability": 0.6904296875
                },
                {
                    "start": 745.76,
                    "end": 745.82,
                    "word": ",",
                    "probability": 0.78955078125
                },
                {
                    "start": 745.82,
                    "end": 745.92,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 745.92,
                    "end": 746.18,
                    "word": " challenge",
                    "probability": 0.78955078125
                },
                {
                    "start": 746.18,
                    "end": 746.32,
                    "word": " to",
                    "probability": 0.896484375
                },
                {
                    "start": 746.32,
                    "end": 747.84,
                    "word": " the",
                    "probability": 0.8134765625
                }
            ]
        },
        {
            "id": 122,
            "seek": 75840,
            "start": 747.84,
            "end": 753.44,
            "text": " left off and write something coherent. And if it can, that reflects something around its ability",
            "tokens": [
                50372,
                1411,
                766,
                293,
                2464,
                746,
                36239,
                13,
                400,
                498,
                309,
                393,
                11,
                300,
                18926,
                746,
                926,
                1080,
                3485,
                50640
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2514349556699091,
            "compression_ratio": 1.6091549295774648,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 747.84,
                    "end": 748.2,
                    "word": " left",
                    "probability": 0.12188720703125
                },
                {
                    "start": 748.2,
                    "end": 748.58,
                    "word": " off",
                    "probability": 0.83154296875
                },
                {
                    "start": 748.58,
                    "end": 749.14,
                    "word": " and",
                    "probability": 0.7353515625
                },
                {
                    "start": 749.14,
                    "end": 749.34,
                    "word": " write",
                    "probability": 0.75537109375
                },
                {
                    "start": 749.34,
                    "end": 749.62,
                    "word": " something",
                    "probability": 0.72021484375
                },
                {
                    "start": 749.62,
                    "end": 750.02,
                    "word": " coherent",
                    "probability": 0.7841796875
                },
                {
                    "start": 750.02,
                    "end": 750.16,
                    "word": ".",
                    "probability": 0.4384765625
                },
                {
                    "start": 750.16,
                    "end": 750.22,
                    "word": " And",
                    "probability": 0.87255859375
                },
                {
                    "start": 750.22,
                    "end": 750.32,
                    "word": " if",
                    "probability": 0.890625
                },
                {
                    "start": 750.32,
                    "end": 750.42,
                    "word": " it",
                    "probability": 0.8984375
                },
                {
                    "start": 750.42,
                    "end": 750.68,
                    "word": " can",
                    "probability": 0.90087890625
                },
                {
                    "start": 750.68,
                    "end": 750.82,
                    "word": ",",
                    "probability": 0.82373046875
                },
                {
                    "start": 750.82,
                    "end": 751.02,
                    "word": " that",
                    "probability": 0.8759765625
                },
                {
                    "start": 751.02,
                    "end": 751.34,
                    "word": " reflects",
                    "probability": 0.73681640625
                },
                {
                    "start": 751.34,
                    "end": 751.72,
                    "word": " something",
                    "probability": 0.7197265625
                },
                {
                    "start": 751.72,
                    "end": 752.56,
                    "word": " around",
                    "probability": 0.8603515625
                },
                {
                    "start": 752.56,
                    "end": 753.14,
                    "word": " its",
                    "probability": 0.7822265625
                },
                {
                    "start": 753.14,
                    "end": 753.44,
                    "word": " ability",
                    "probability": 0.86669921875
                }
            ]
        },
        {
            "id": 123,
            "seek": 75840,
            "start": 753.44,
            "end": 760.74,
            "text": " to solve problems like the fundamental problems I described earlier. So the neural network",
            "tokens": [
                50640,
                281,
                5039,
                2740,
                411,
                264,
                8088,
                2740,
                286,
                7619,
                3071,
                13,
                407,
                264,
                18161,
                3209,
                51008
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2514349556699091,
            "compression_ratio": 1.6091549295774648,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 753.44,
                    "end": 753.62,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 753.62,
                    "end": 753.82,
                    "word": " solve",
                    "probability": 0.84228515625
                },
                {
                    "start": 753.82,
                    "end": 754.22,
                    "word": " problems",
                    "probability": 0.69140625
                },
                {
                    "start": 754.22,
                    "end": 754.36,
                    "word": " like",
                    "probability": 0.7802734375
                },
                {
                    "start": 754.36,
                    "end": 754.48,
                    "word": " the",
                    "probability": 0.81689453125
                },
                {
                    "start": 754.48,
                    "end": 754.72,
                    "word": " fundamental",
                    "probability": 0.7412109375
                },
                {
                    "start": 754.72,
                    "end": 755.16,
                    "word": " problems",
                    "probability": 0.69189453125
                },
                {
                    "start": 755.16,
                    "end": 755.38,
                    "word": " I",
                    "probability": 0.927734375
                },
                {
                    "start": 755.38,
                    "end": 756.5,
                    "word": " described",
                    "probability": 0.66943359375
                },
                {
                    "start": 756.5,
                    "end": 756.98,
                    "word": " earlier",
                    "probability": 0.92236328125
                },
                {
                    "start": 756.98,
                    "end": 757.54,
                    "word": ".",
                    "probability": 0.9560546875
                },
                {
                    "start": 757.54,
                    "end": 757.78,
                    "word": " So",
                    "probability": 0.82763671875
                },
                {
                    "start": 757.78,
                    "end": 758.28,
                    "word": " the",
                    "probability": 0.681640625
                },
                {
                    "start": 758.28,
                    "end": 760.48,
                    "word": " neural",
                    "probability": 0.703125
                },
                {
                    "start": 760.48,
                    "end": 760.74,
                    "word": " network",
                    "probability": 0.82763671875
                }
            ]
        },
        {
            "id": 124,
            "seek": 75840,
            "start": 760.74,
            "end": 765.98,
            "text": " continues. The scientists named the population after their distinctive horn, Ovid's unicorn.",
            "tokens": [
                51008,
                6515,
                13,
                440,
                7708,
                4926,
                264,
                4415,
                934,
                641,
                27766,
                13482,
                11,
                422,
                6833,
                311,
                28122,
                13,
                51264
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2514349556699091,
            "compression_ratio": 1.6091549295774648,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 760.74,
                    "end": 761.18,
                    "word": " continues",
                    "probability": 0.7998046875
                },
                {
                    "start": 761.18,
                    "end": 761.4,
                    "word": ".",
                    "probability": 0.55126953125
                },
                {
                    "start": 761.4,
                    "end": 761.6,
                    "word": " The",
                    "probability": 0.5576171875
                },
                {
                    "start": 761.6,
                    "end": 762.0,
                    "word": " scientists",
                    "probability": 0.421142578125
                },
                {
                    "start": 762.0,
                    "end": 762.28,
                    "word": " named",
                    "probability": 0.75732421875
                },
                {
                    "start": 762.28,
                    "end": 762.42,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 762.42,
                    "end": 762.88,
                    "word": " population",
                    "probability": 0.8623046875
                },
                {
                    "start": 762.88,
                    "end": 763.18,
                    "word": " after",
                    "probability": 0.701171875
                },
                {
                    "start": 763.18,
                    "end": 763.36,
                    "word": " their",
                    "probability": 0.7568359375
                },
                {
                    "start": 763.36,
                    "end": 763.68,
                    "word": " distinctive",
                    "probability": 0.87548828125
                },
                {
                    "start": 763.68,
                    "end": 764.14,
                    "word": " horn",
                    "probability": 0.81396484375
                },
                {
                    "start": 764.14,
                    "end": 764.74,
                    "word": ",",
                    "probability": 0.82421875
                },
                {
                    "start": 764.74,
                    "end": 765.08,
                    "word": " Ovid's",
                    "probability": 0.8243815104166666
                },
                {
                    "start": 765.08,
                    "end": 765.4,
                    "word": " unicorn",
                    "probability": 0.59765625
                },
                {
                    "start": 765.4,
                    "end": 765.98,
                    "word": ".",
                    "probability": 0.970703125
                }
            ]
        },
        {
            "id": 125,
            "seek": 75840,
            "start": 765.98,
            "end": 770.18,
            "text": " These four-horned, silver-white unicorns were previously unknown to science.",
            "tokens": [
                51264,
                1981,
                1451,
                12,
                31990,
                292,
                11,
                8753,
                12,
                28865,
                28122,
                82,
                645,
                8046,
                9841,
                281,
                3497,
                13,
                51452
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2514349556699091,
            "compression_ratio": 1.6091549295774648,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 765.98,
                    "end": 766.18,
                    "word": " These",
                    "probability": 0.904296875
                },
                {
                    "start": 766.18,
                    "end": 766.48,
                    "word": " four",
                    "probability": 0.85107421875
                },
                {
                    "start": 766.48,
                    "end": 766.88,
                    "word": "-horned",
                    "probability": 0.6766764322916666
                },
                {
                    "start": 766.88,
                    "end": 766.9,
                    "word": ",",
                    "probability": 0.3212890625
                },
                {
                    "start": 766.9,
                    "end": 767.12,
                    "word": " silver",
                    "probability": 0.845703125
                },
                {
                    "start": 767.12,
                    "end": 767.46,
                    "word": "-white",
                    "probability": 0.81689453125
                },
                {
                    "start": 767.46,
                    "end": 768.08,
                    "word": " unicorns",
                    "probability": 0.833740234375
                },
                {
                    "start": 768.08,
                    "end": 768.2,
                    "word": " were",
                    "probability": 0.87451171875
                },
                {
                    "start": 768.2,
                    "end": 768.6,
                    "word": " previously",
                    "probability": 0.7587890625
                },
                {
                    "start": 768.6,
                    "end": 769.0,
                    "word": " unknown",
                    "probability": 0.759765625
                },
                {
                    "start": 769.0,
                    "end": 769.24,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 769.24,
                    "end": 769.7,
                    "word": " science",
                    "probability": 0.8837890625
                },
                {
                    "start": 769.7,
                    "end": 770.18,
                    "word": ".",
                    "probability": 0.9951171875
                }
            ]
        },
        {
            "id": 126,
            "seek": 75840,
            "start": 770.18,
            "end": 775.38,
            "text": " Now, after almost two centuries, the mystery of what sparked this odd phenomenon is finally solved.",
            "tokens": [
                51492,
                823,
                11,
                934,
                1920,
                732,
                13926,
                11,
                264,
                11422,
                295,
                437,
                39653,
                341,
                7401,
                14029,
                307,
                2721,
                13041,
                13,
                51736
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2514349556699091,
            "compression_ratio": 1.6091549295774648,
            "no_speech_prob": 0.00048732757568359375,
            "words": [
                {
                    "start": 770.18,
                    "end": 770.74,
                    "word": " Now",
                    "probability": 0.88671875
                },
                {
                    "start": 770.74,
                    "end": 770.88,
                    "word": ",",
                    "probability": 0.75927734375
                },
                {
                    "start": 770.88,
                    "end": 771.06,
                    "word": " after",
                    "probability": 0.7392578125
                },
                {
                    "start": 771.06,
                    "end": 771.36,
                    "word": " almost",
                    "probability": 0.69287109375
                },
                {
                    "start": 771.36,
                    "end": 771.62,
                    "word": " two",
                    "probability": 0.81494140625
                },
                {
                    "start": 771.62,
                    "end": 772.0,
                    "word": " centuries",
                    "probability": 0.822265625
                },
                {
                    "start": 772.0,
                    "end": 772.16,
                    "word": ",",
                    "probability": 0.98193359375
                },
                {
                    "start": 772.16,
                    "end": 772.3,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 772.3,
                    "end": 772.56,
                    "word": " mystery",
                    "probability": 0.76904296875
                },
                {
                    "start": 772.56,
                    "end": 772.72,
                    "word": " of",
                    "probability": 0.89453125
                },
                {
                    "start": 772.72,
                    "end": 772.84,
                    "word": " what",
                    "probability": 0.87841796875
                },
                {
                    "start": 772.84,
                    "end": 773.1,
                    "word": " sparked",
                    "probability": 0.80224609375
                },
                {
                    "start": 773.1,
                    "end": 773.32,
                    "word": " this",
                    "probability": 0.8720703125
                },
                {
                    "start": 773.32,
                    "end": 773.56,
                    "word": " odd",
                    "probability": 0.71142578125
                },
                {
                    "start": 773.56,
                    "end": 773.94,
                    "word": " phenomenon",
                    "probability": 0.7890625
                },
                {
                    "start": 773.94,
                    "end": 774.16,
                    "word": " is",
                    "probability": 0.873046875
                },
                {
                    "start": 774.16,
                    "end": 774.46,
                    "word": " finally",
                    "probability": 0.8193359375
                },
                {
                    "start": 774.46,
                    "end": 774.94,
                    "word": " solved",
                    "probability": 0.78515625
                },
                {
                    "start": 774.94,
                    "end": 775.38,
                    "word": ".",
                    "probability": 0.99169921875
                }
            ]
        },
        {
            "id": 127,
            "seek": 78272,
            "start": 775.38,
            "end": 780.44,
            "text": " Dr. Jorge Perez, an evolutionary biologist from the University of La Paz and several companions,",
            "tokens": [
                50364,
                2491,
                13,
                36875,
                47317,
                11,
                364,
                27567,
                3228,
                9201,
                490,
                264,
                3535,
                295,
                2369,
                430,
                921,
                293,
                2940,
                28009,
                11,
                50612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19566441978420224,
            "compression_ratio": 1.6597938144329898,
            "no_speech_prob": 0.0009393692016601562,
            "words": [
                {
                    "start": 775.38,
                    "end": 775.64,
                    "word": " Dr",
                    "probability": 0.378173828125
                },
                {
                    "start": 775.64,
                    "end": 775.74,
                    "word": ".",
                    "probability": 0.9248046875
                },
                {
                    "start": 775.74,
                    "end": 776.0,
                    "word": " Jorge",
                    "probability": 0.91650390625
                },
                {
                    "start": 776.0,
                    "end": 776.42,
                    "word": " Perez",
                    "probability": 0.75830078125
                },
                {
                    "start": 776.42,
                    "end": 776.58,
                    "word": ",",
                    "probability": 0.89501953125
                },
                {
                    "start": 776.58,
                    "end": 776.72,
                    "word": " an",
                    "probability": 0.89111328125
                },
                {
                    "start": 776.72,
                    "end": 777.12,
                    "word": " evolutionary",
                    "probability": 0.7490234375
                },
                {
                    "start": 777.12,
                    "end": 777.7,
                    "word": " biologist",
                    "probability": 0.91162109375
                },
                {
                    "start": 777.7,
                    "end": 777.98,
                    "word": " from",
                    "probability": 0.81396484375
                },
                {
                    "start": 777.98,
                    "end": 778.12,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 778.12,
                    "end": 778.48,
                    "word": " University",
                    "probability": 0.65478515625
                },
                {
                    "start": 778.48,
                    "end": 778.68,
                    "word": " of",
                    "probability": 0.89404296875
                },
                {
                    "start": 778.68,
                    "end": 778.82,
                    "word": " La",
                    "probability": 0.87060546875
                },
                {
                    "start": 778.82,
                    "end": 779.08,
                    "word": " Paz",
                    "probability": 0.966552734375
                },
                {
                    "start": 779.08,
                    "end": 779.26,
                    "word": " and",
                    "probability": 0.69873046875
                },
                {
                    "start": 779.26,
                    "end": 779.48,
                    "word": " several",
                    "probability": 0.8271484375
                },
                {
                    "start": 779.48,
                    "end": 779.96,
                    "word": " companions",
                    "probability": 0.75927734375
                },
                {
                    "start": 779.96,
                    "end": 780.44,
                    "word": ",",
                    "probability": 0.55029296875
                }
            ]
        },
        {
            "id": 128,
            "seek": 78272,
            "start": 780.44,
            "end": 785.46,
            "text": " were exploring the Andes Mountains when they found a small valley with no other animals or humans.",
            "tokens": [
                50612,
                645,
                12736,
                264,
                400,
                279,
                30970,
                562,
                436,
                1352,
                257,
                1359,
                17636,
                365,
                572,
                661,
                4882,
                420,
                6255,
                13,
                50844
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19566441978420224,
            "compression_ratio": 1.6597938144329898,
            "no_speech_prob": 0.0009393692016601562,
            "words": [
                {
                    "start": 780.44,
                    "end": 780.6,
                    "word": " were",
                    "probability": 0.8447265625
                },
                {
                    "start": 780.6,
                    "end": 780.98,
                    "word": " exploring",
                    "probability": 0.8095703125
                },
                {
                    "start": 780.98,
                    "end": 781.16,
                    "word": " the",
                    "probability": 0.79296875
                },
                {
                    "start": 781.16,
                    "end": 781.42,
                    "word": " Andes",
                    "probability": 0.938720703125
                },
                {
                    "start": 781.42,
                    "end": 781.78,
                    "word": " Mountains",
                    "probability": 0.439208984375
                },
                {
                    "start": 781.78,
                    "end": 782.06,
                    "word": " when",
                    "probability": 0.85595703125
                },
                {
                    "start": 782.06,
                    "end": 782.22,
                    "word": " they",
                    "probability": 0.7294921875
                },
                {
                    "start": 782.22,
                    "end": 782.46,
                    "word": " found",
                    "probability": 0.88671875
                },
                {
                    "start": 782.62,
                    "end": 782.98,
                    "word": " a small",
                    "probability": 0.8486328125
                },
                {
                    "start": 782.98,
                    "end": 783.36,
                    "word": " valley",
                    "probability": 0.7529296875
                },
                {
                    "start": 783.36,
                    "end": 783.68,
                    "word": " with",
                    "probability": 0.77978515625
                },
                {
                    "start": 783.68,
                    "end": 783.88,
                    "word": " no",
                    "probability": 0.90673828125
                },
                {
                    "start": 783.88,
                    "end": 784.08,
                    "word": " other",
                    "probability": 0.78076171875
                },
                {
                    "start": 784.08,
                    "end": 784.42,
                    "word": " animals",
                    "probability": 0.8193359375
                },
                {
                    "start": 784.42,
                    "end": 784.6,
                    "word": " or",
                    "probability": 0.90283203125
                },
                {
                    "start": 784.6,
                    "end": 784.98,
                    "word": " humans",
                    "probability": 0.9169921875
                },
                {
                    "start": 784.98,
                    "end": 785.46,
                    "word": ".",
                    "probability": 0.9755859375
                }
            ]
        },
        {
            "id": 129,
            "seek": 78272,
            "start": 785.86,
            "end": 789.86,
            "text": " Perez noticed that the valley had what appeared to be a natural fountain, surrounded by two peaks",
            "tokens": [
                50872,
                47317,
                5694,
                300,
                264,
                17636,
                632,
                437,
                8516,
                281,
                312,
                257,
                3303,
                29451,
                11,
                13221,
                538,
                732,
                26897,
                51092
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19566441978420224,
            "compression_ratio": 1.6597938144329898,
            "no_speech_prob": 0.0009393692016601562,
            "words": [
                {
                    "start": 785.86,
                    "end": 785.86,
                    "word": " Perez",
                    "probability": 0.8662109375
                },
                {
                    "start": 785.86,
                    "end": 786.2,
                    "word": " noticed",
                    "probability": 0.85400390625
                },
                {
                    "start": 786.2,
                    "end": 786.38,
                    "word": " that",
                    "probability": 0.87841796875
                },
                {
                    "start": 786.38,
                    "end": 786.52,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 786.52,
                    "end": 786.72,
                    "word": " valley",
                    "probability": 0.7822265625
                },
                {
                    "start": 786.72,
                    "end": 786.9,
                    "word": " had",
                    "probability": 0.8076171875
                },
                {
                    "start": 786.9,
                    "end": 787.04,
                    "word": " what",
                    "probability": 0.87158203125
                },
                {
                    "start": 787.04,
                    "end": 787.32,
                    "word": " appeared",
                    "probability": 0.7578125
                },
                {
                    "start": 787.32,
                    "end": 787.48,
                    "word": " to",
                    "probability": 0.9072265625
                },
                {
                    "start": 787.48,
                    "end": 787.6,
                    "word": " be",
                    "probability": 0.9072265625
                },
                {
                    "start": 787.72,
                    "end": 788.04,
                    "word": " a natural",
                    "probability": 0.84375
                },
                {
                    "start": 788.04,
                    "end": 788.52,
                    "word": " fountain",
                    "probability": 0.74072265625
                },
                {
                    "start": 788.52,
                    "end": 788.88,
                    "word": ",",
                    "probability": 0.371826171875
                },
                {
                    "start": 788.88,
                    "end": 789.08,
                    "word": " surrounded",
                    "probability": 0.720703125
                },
                {
                    "start": 789.08,
                    "end": 789.34,
                    "word": " by",
                    "probability": 0.9072265625
                },
                {
                    "start": 789.34,
                    "end": 789.58,
                    "word": " two",
                    "probability": 0.82421875
                },
                {
                    "start": 789.58,
                    "end": 789.86,
                    "word": " peaks",
                    "probability": 0.81982421875
                }
            ]
        },
        {
            "id": 130,
            "seek": 78272,
            "start": 789.86,
            "end": 795.36,
            "text": " of rock and silver snow. Perez and the others then ventured further into the valley. By the time we",
            "tokens": [
                51092,
                295,
                3727,
                293,
                8753,
                5756,
                13,
                47317,
                293,
                264,
                2357,
                550,
                6931,
                3831,
                3052,
                666,
                264,
                17636,
                13,
                3146,
                264,
                565,
                321,
                51360
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19566441978420224,
            "compression_ratio": 1.6597938144329898,
            "no_speech_prob": 0.0009393692016601562,
            "words": [
                {
                    "start": 789.86,
                    "end": 790.1,
                    "word": " of",
                    "probability": 0.90869140625
                },
                {
                    "start": 790.1,
                    "end": 790.42,
                    "word": " rock",
                    "probability": 0.8330078125
                },
                {
                    "start": 790.42,
                    "end": 790.6,
                    "word": " and",
                    "probability": 0.88916015625
                },
                {
                    "start": 790.6,
                    "end": 790.88,
                    "word": " silver",
                    "probability": 0.83056640625
                },
                {
                    "start": 790.88,
                    "end": 791.34,
                    "word": " snow",
                    "probability": 0.8466796875
                },
                {
                    "start": 791.34,
                    "end": 791.82,
                    "word": ".",
                    "probability": 0.98828125
                },
                {
                    "start": 792.3,
                    "end": 792.3,
                    "word": " Perez",
                    "probability": 0.8427734375
                },
                {
                    "start": 792.3,
                    "end": 792.5,
                    "word": " and",
                    "probability": 0.8935546875
                },
                {
                    "start": 792.5,
                    "end": 792.6,
                    "word": " the",
                    "probability": 0.80224609375
                },
                {
                    "start": 792.6,
                    "end": 792.84,
                    "word": " others",
                    "probability": 0.7685546875
                },
                {
                    "start": 792.84,
                    "end": 793.06,
                    "word": " then",
                    "probability": 0.69873046875
                },
                {
                    "start": 793.06,
                    "end": 793.46,
                    "word": " ventured",
                    "probability": 0.87646484375
                },
                {
                    "start": 793.46,
                    "end": 793.84,
                    "word": " further",
                    "probability": 0.85107421875
                },
                {
                    "start": 793.84,
                    "end": 794.04,
                    "word": " into",
                    "probability": 0.7255859375
                },
                {
                    "start": 794.04,
                    "end": 794.22,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 794.22,
                    "end": 794.48,
                    "word": " valley",
                    "probability": 0.7744140625
                },
                {
                    "start": 794.48,
                    "end": 794.94,
                    "word": ".",
                    "probability": 0.93896484375
                },
                {
                    "start": 794.94,
                    "end": 794.98,
                    "word": " By",
                    "probability": 0.498046875
                },
                {
                    "start": 794.98,
                    "end": 795.12,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 795.12,
                    "end": 795.26,
                    "word": " time",
                    "probability": 0.806640625
                },
                {
                    "start": 795.26,
                    "end": 795.36,
                    "word": " we",
                    "probability": 0.82421875
                }
            ]
        },
        {
            "id": 131,
            "seek": 78272,
            "start": 795.36,
            "end": 799.7,
            "text": " reached the top of one peak, the water looked blue with some crystals on top, said Perez.",
            "tokens": [
                51360,
                6488,
                264,
                1192,
                295,
                472,
                10651,
                11,
                264,
                1281,
                2956,
                3344,
                365,
                512,
                23772,
                322,
                1192,
                11,
                848,
                47317,
                13,
                51556
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19566441978420224,
            "compression_ratio": 1.6597938144329898,
            "no_speech_prob": 0.0009393692016601562,
            "words": [
                {
                    "start": 795.36,
                    "end": 795.54,
                    "word": " reached",
                    "probability": 0.7158203125
                },
                {
                    "start": 795.54,
                    "end": 795.7,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 795.7,
                    "end": 795.88,
                    "word": " top",
                    "probability": 0.8203125
                },
                {
                    "start": 795.88,
                    "end": 796.0,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 796.0,
                    "end": 796.2,
                    "word": " one",
                    "probability": 0.81201171875
                },
                {
                    "start": 796.2,
                    "end": 796.48,
                    "word": " peak",
                    "probability": 0.818359375
                },
                {
                    "start": 796.48,
                    "end": 796.58,
                    "word": ",",
                    "probability": 0.93408203125
                },
                {
                    "start": 796.58,
                    "end": 796.68,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 796.68,
                    "end": 796.84,
                    "word": " water",
                    "probability": 0.79296875
                },
                {
                    "start": 796.84,
                    "end": 797.08,
                    "word": " looked",
                    "probability": 0.71533203125
                },
                {
                    "start": 797.08,
                    "end": 797.36,
                    "word": " blue",
                    "probability": 0.9443359375
                },
                {
                    "start": 797.36,
                    "end": 797.52,
                    "word": " with",
                    "probability": 0.7392578125
                },
                {
                    "start": 797.52,
                    "end": 797.7,
                    "word": " some",
                    "probability": 0.79931640625
                },
                {
                    "start": 797.7,
                    "end": 798.0,
                    "word": " crystals",
                    "probability": 0.7744140625
                },
                {
                    "start": 798.0,
                    "end": 798.26,
                    "word": " on",
                    "probability": 0.900390625
                },
                {
                    "start": 798.26,
                    "end": 798.6,
                    "word": " top",
                    "probability": 0.74560546875
                },
                {
                    "start": 798.6,
                    "end": 798.64,
                    "word": ",",
                    "probability": 0.77587890625
                },
                {
                    "start": 798.64,
                    "end": 798.8,
                    "word": " said",
                    "probability": 0.8427734375
                },
                {
                    "start": 798.8,
                    "end": 799.22,
                    "word": " Perez",
                    "probability": 0.83935546875
                },
                {
                    "start": 799.22,
                    "end": 799.7,
                    "word": ".",
                    "probability": 0.568359375
                }
            ]
        },
        {
            "id": 132,
            "seek": 80824,
            "start": 799.7,
            "end": 807.28,
            "text": " And the essay continues. But what interested Younghoo and I in seeing examples of what",
            "tokens": [
                50372,
                400,
                264,
                16238,
                6515,
                13,
                583,
                437,
                3102,
                8160,
                19069,
                293,
                286,
                294,
                2577,
                5110,
                295,
                437,
                50736
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2527689805513696,
            "compression_ratio": 1.7464114832535884,
            "no_speech_prob": 0.051727294921875,
            "words": [
                {
                    "start": 799.7,
                    "end": 800.16,
                    "word": " And",
                    "probability": 0.321044921875
                },
                {
                    "start": 800.16,
                    "end": 800.52,
                    "word": " the",
                    "probability": 0.51611328125
                },
                {
                    "start": 800.52,
                    "end": 800.66,
                    "word": " essay",
                    "probability": 0.73095703125
                },
                {
                    "start": 800.66,
                    "end": 801.08,
                    "word": " continues",
                    "probability": 0.7958984375
                },
                {
                    "start": 801.08,
                    "end": 801.92,
                    "word": ".",
                    "probability": 0.86279296875
                },
                {
                    "start": 802.82,
                    "end": 803.3,
                    "word": " But",
                    "probability": 0.72900390625
                },
                {
                    "start": 803.3,
                    "end": 804.0,
                    "word": " what",
                    "probability": 0.84375
                },
                {
                    "start": 804.0,
                    "end": 804.42,
                    "word": " interested",
                    "probability": 0.63720703125
                },
                {
                    "start": 804.42,
                    "end": 805.2,
                    "word": " Younghoo",
                    "probability": 0.2843017578125
                },
                {
                    "start": 805.2,
                    "end": 805.34,
                    "word": " and",
                    "probability": 0.89013671875
                },
                {
                    "start": 805.34,
                    "end": 805.54,
                    "word": " I",
                    "probability": 0.9970703125
                },
                {
                    "start": 805.54,
                    "end": 806.02,
                    "word": " in",
                    "probability": 0.703125
                },
                {
                    "start": 806.02,
                    "end": 806.56,
                    "word": " seeing",
                    "probability": 0.736328125
                },
                {
                    "start": 806.56,
                    "end": 807.02,
                    "word": " examples",
                    "probability": 0.8076171875
                },
                {
                    "start": 807.02,
                    "end": 807.14,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 807.14,
                    "end": 807.28,
                    "word": " what",
                    "probability": 0.8701171875
                }
            ]
        },
        {
            "id": 133,
            "seek": 80824,
            "start": 807.28,
            "end": 812.72,
            "text": " transformers can do like this is that it seems that the transformers have, examples like these",
            "tokens": [
                50736,
                4088,
                433,
                393,
                360,
                411,
                341,
                307,
                300,
                309,
                2544,
                300,
                264,
                4088,
                433,
                362,
                11,
                5110,
                411,
                613,
                51024
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2527689805513696,
            "compression_ratio": 1.7464114832535884,
            "no_speech_prob": 0.051727294921875,
            "words": [
                {
                    "start": 807.28,
                    "end": 807.74,
                    "word": " transformers",
                    "probability": 0.8447265625
                },
                {
                    "start": 807.74,
                    "end": 807.88,
                    "word": " can",
                    "probability": 0.89892578125
                },
                {
                    "start": 807.88,
                    "end": 808.04,
                    "word": " do",
                    "probability": 0.9091796875
                },
                {
                    "start": 808.04,
                    "end": 808.22,
                    "word": " like",
                    "probability": 0.8779296875
                },
                {
                    "start": 808.22,
                    "end": 808.48,
                    "word": " this",
                    "probability": 0.88134765625
                },
                {
                    "start": 808.48,
                    "end": 808.68,
                    "word": " is",
                    "probability": 0.724609375
                },
                {
                    "start": 808.68,
                    "end": 809.02,
                    "word": " that",
                    "probability": 0.869140625
                },
                {
                    "start": 809.02,
                    "end": 809.58,
                    "word": " it",
                    "probability": 0.8212890625
                },
                {
                    "start": 809.58,
                    "end": 809.82,
                    "word": " seems",
                    "probability": 0.6650390625
                },
                {
                    "start": 809.82,
                    "end": 810.0,
                    "word": " that",
                    "probability": 0.865234375
                },
                {
                    "start": 810.0,
                    "end": 810.12,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 810.12,
                    "end": 810.92,
                    "word": " transformers",
                    "probability": 0.908203125
                },
                {
                    "start": 810.92,
                    "end": 811.46,
                    "word": " have",
                    "probability": 0.87548828125
                },
                {
                    "start": 811.46,
                    "end": 811.82,
                    "word": ",",
                    "probability": 0.392333984375
                },
                {
                    "start": 811.82,
                    "end": 812.26,
                    "word": " examples",
                    "probability": 0.74951171875
                },
                {
                    "start": 812.26,
                    "end": 812.44,
                    "word": " like",
                    "probability": 0.873046875
                },
                {
                    "start": 812.44,
                    "end": 812.72,
                    "word": " these",
                    "probability": 0.65869140625
                }
            ]
        },
        {
            "id": 134,
            "seek": 80824,
            "start": 812.72,
            "end": 817.68,
            "text": " suggest that transformers have understood the syntax and semantics of the English language",
            "tokens": [
                51024,
                3402,
                300,
                4088,
                433,
                362,
                7320,
                264,
                28431,
                293,
                4361,
                45298,
                295,
                264,
                3669,
                2856,
                51276
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2527689805513696,
            "compression_ratio": 1.7464114832535884,
            "no_speech_prob": 0.051727294921875,
            "words": [
                {
                    "start": 812.72,
                    "end": 813.36,
                    "word": " suggest",
                    "probability": 0.58642578125
                },
                {
                    "start": 813.36,
                    "end": 813.54,
                    "word": " that",
                    "probability": 0.45361328125
                },
                {
                    "start": 813.54,
                    "end": 813.98,
                    "word": " transformers",
                    "probability": 0.70947265625
                },
                {
                    "start": 813.98,
                    "end": 814.12,
                    "word": " have",
                    "probability": 0.87841796875
                },
                {
                    "start": 814.12,
                    "end": 814.46,
                    "word": " understood",
                    "probability": 0.8349609375
                },
                {
                    "start": 814.46,
                    "end": 814.74,
                    "word": " the",
                    "probability": 0.78564453125
                },
                {
                    "start": 814.74,
                    "end": 815.18,
                    "word": " syntax",
                    "probability": 0.91162109375
                },
                {
                    "start": 815.18,
                    "end": 815.86,
                    "word": " and",
                    "probability": 0.89013671875
                },
                {
                    "start": 815.86,
                    "end": 816.3,
                    "word": " semantics",
                    "probability": 0.953125
                },
                {
                    "start": 816.3,
                    "end": 816.76,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 816.76,
                    "end": 816.98,
                    "word": " the",
                    "probability": 0.828125
                },
                {
                    "start": 816.98,
                    "end": 817.2,
                    "word": " English",
                    "probability": 0.8984375
                },
                {
                    "start": 817.2,
                    "end": 817.68,
                    "word": " language",
                    "probability": 0.6865234375
                }
            ]
        },
        {
            "id": 135,
            "seek": 80824,
            "start": 817.68,
            "end": 825.22,
            "text": " to some degree, and which begs the question, could the representations that the transformers",
            "tokens": [
                51276,
                281,
                512,
                4314,
                11,
                293,
                597,
                4612,
                82,
                264,
                1168,
                11,
                727,
                264,
                33358,
                300,
                264,
                4088,
                433,
                51640
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2527689805513696,
            "compression_ratio": 1.7464114832535884,
            "no_speech_prob": 0.051727294921875,
            "words": [
                {
                    "start": 817.68,
                    "end": 818.2,
                    "word": " to",
                    "probability": 0.85791015625
                },
                {
                    "start": 818.2,
                    "end": 818.36,
                    "word": " some",
                    "probability": 0.81689453125
                },
                {
                    "start": 818.36,
                    "end": 818.68,
                    "word": " degree",
                    "probability": 0.87841796875
                },
                {
                    "start": 818.68,
                    "end": 818.96,
                    "word": ",",
                    "probability": 0.54541015625
                },
                {
                    "start": 818.96,
                    "end": 819.02,
                    "word": " and",
                    "probability": 0.3837890625
                },
                {
                    "start": 819.02,
                    "end": 820.18,
                    "word": " which",
                    "probability": 0.82666015625
                },
                {
                    "start": 820.18,
                    "end": 820.42,
                    "word": " begs",
                    "probability": 0.85498046875
                },
                {
                    "start": 820.42,
                    "end": 820.54,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 820.54,
                    "end": 820.98,
                    "word": " question",
                    "probability": 0.80419921875
                },
                {
                    "start": 820.98,
                    "end": 821.68,
                    "word": ",",
                    "probability": 0.82421875
                },
                {
                    "start": 821.68,
                    "end": 822.16,
                    "word": " could",
                    "probability": 0.7587890625
                },
                {
                    "start": 822.16,
                    "end": 822.66,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 822.66,
                    "end": 823.24,
                    "word": " representations",
                    "probability": 0.81689453125
                },
                {
                    "start": 823.24,
                    "end": 824.02,
                    "word": " that",
                    "probability": 0.8828125
                },
                {
                    "start": 824.02,
                    "end": 824.76,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 824.76,
                    "end": 825.22,
                    "word": " transformers",
                    "probability": 0.931396484375
                }
            ]
        },
        {
            "id": 136,
            "seek": 83760,
            "start": 825.22,
            "end": 832.14,
            "text": " learn in their parameter structure be useful in detecting phishing emails, since these",
            "tokens": [
                50364,
                1466,
                294,
                641,
                13075,
                3877,
                312,
                4420,
                294,
                40237,
                903,
                3807,
                12524,
                11,
                1670,
                613,
                50708
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26874999940395355,
            "compression_ratio": 1.6753731343283582,
            "no_speech_prob": 0.00628662109375,
            "words": [
                {
                    "start": 825.22,
                    "end": 825.54,
                    "word": " learn",
                    "probability": 0.634765625
                },
                {
                    "start": 825.54,
                    "end": 825.78,
                    "word": " in",
                    "probability": 0.6611328125
                },
                {
                    "start": 825.78,
                    "end": 826.0,
                    "word": " their",
                    "probability": 0.89404296875
                },
                {
                    "start": 826.0,
                    "end": 827.16,
                    "word": " parameter",
                    "probability": 0.6650390625
                },
                {
                    "start": 827.16,
                    "end": 827.96,
                    "word": " structure",
                    "probability": 0.708984375
                },
                {
                    "start": 827.96,
                    "end": 829.24,
                    "word": " be",
                    "probability": 0.65966796875
                },
                {
                    "start": 829.24,
                    "end": 829.58,
                    "word": " useful",
                    "probability": 0.78076171875
                },
                {
                    "start": 829.58,
                    "end": 830.02,
                    "word": " in",
                    "probability": 0.8935546875
                },
                {
                    "start": 830.02,
                    "end": 830.44,
                    "word": " detecting",
                    "probability": 0.779296875
                },
                {
                    "start": 830.44,
                    "end": 830.8,
                    "word": " phishing",
                    "probability": 0.8369140625
                },
                {
                    "start": 830.8,
                    "end": 831.08,
                    "word": " emails",
                    "probability": 0.71533203125
                },
                {
                    "start": 831.08,
                    "end": 831.3,
                    "word": ",",
                    "probability": 0.327880859375
                },
                {
                    "start": 831.3,
                    "end": 831.5,
                    "word": " since",
                    "probability": 0.7861328125
                },
                {
                    "start": 831.5,
                    "end": 832.14,
                    "word": " these",
                    "probability": 0.61328125
                }
            ]
        },
        {
            "id": 137,
            "seek": 83760,
            "start": 832.14,
            "end": 839.08,
            "text": " transformer models seem to have distilled knowledge of how language works and what it means",
            "tokens": [
                50708,
                31782,
                5245,
                1643,
                281,
                362,
                1483,
                6261,
                3601,
                295,
                577,
                2856,
                1985,
                293,
                437,
                309,
                1355,
                51048
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26874999940395355,
            "compression_ratio": 1.6753731343283582,
            "no_speech_prob": 0.00628662109375,
            "words": [
                {
                    "start": 832.14,
                    "end": 832.54,
                    "word": " transformer",
                    "probability": 0.74853515625
                },
                {
                    "start": 832.54,
                    "end": 832.92,
                    "word": " models",
                    "probability": 0.80712890625
                },
                {
                    "start": 832.92,
                    "end": 833.16,
                    "word": " seem",
                    "probability": 0.85107421875
                },
                {
                    "start": 833.16,
                    "end": 833.56,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 833.56,
                    "end": 834.5,
                    "word": " have",
                    "probability": 0.88232421875
                },
                {
                    "start": 834.5,
                    "end": 836.06,
                    "word": " distilled",
                    "probability": 0.900146484375
                },
                {
                    "start": 836.06,
                    "end": 836.88,
                    "word": " knowledge",
                    "probability": 0.82763671875
                },
                {
                    "start": 836.88,
                    "end": 837.04,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 837.04,
                    "end": 837.2,
                    "word": " how",
                    "probability": 0.89453125
                },
                {
                    "start": 837.2,
                    "end": 837.46,
                    "word": " language",
                    "probability": 0.65576171875
                },
                {
                    "start": 837.46,
                    "end": 837.78,
                    "word": " works",
                    "probability": 0.71728515625
                },
                {
                    "start": 837.78,
                    "end": 838.1,
                    "word": " and",
                    "probability": 0.82763671875
                },
                {
                    "start": 838.1,
                    "end": 838.56,
                    "word": " what",
                    "probability": 0.86083984375
                },
                {
                    "start": 838.56,
                    "end": 838.7,
                    "word": " it",
                    "probability": 0.9033203125
                },
                {
                    "start": 838.7,
                    "end": 839.08,
                    "word": " means",
                    "probability": 0.8759765625
                }
            ]
        },
        {
            "id": 138,
            "seek": 83760,
            "start": 840.6,
            "end": 846.22,
            "text": " within their parameter settings. Okay, so this is one example of, sort of impressive example,",
            "tokens": [
                51152,
                1951,
                641,
                13075,
                6257,
                13,
                1033,
                11,
                370,
                341,
                307,
                472,
                1365,
                295,
                11,
                1333,
                295,
                8992,
                1365,
                11,
                51408
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26874999940395355,
            "compression_ratio": 1.6753731343283582,
            "no_speech_prob": 0.00628662109375,
            "words": [
                {
                    "start": 840.6,
                    "end": 841.16,
                    "word": " within",
                    "probability": 0.734375
                },
                {
                    "start": 841.16,
                    "end": 841.72,
                    "word": " their",
                    "probability": 0.90234375
                },
                {
                    "start": 841.72,
                    "end": 842.04,
                    "word": " parameter",
                    "probability": 0.82421875
                },
                {
                    "start": 842.04,
                    "end": 842.54,
                    "word": " settings",
                    "probability": 0.8046875
                },
                {
                    "start": 842.54,
                    "end": 843.1,
                    "word": ".",
                    "probability": 0.9208984375
                },
                {
                    "start": 843.72,
                    "end": 843.88,
                    "word": " Okay",
                    "probability": 0.466552734375
                },
                {
                    "start": 843.88,
                    "end": 844.0,
                    "word": ",",
                    "probability": 0.69873046875
                },
                {
                    "start": 844.0,
                    "end": 844.1,
                    "word": " so",
                    "probability": 0.8896484375
                },
                {
                    "start": 844.1,
                    "end": 844.22,
                    "word": " this",
                    "probability": 0.85791015625
                },
                {
                    "start": 844.22,
                    "end": 844.28,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 844.28,
                    "end": 844.4,
                    "word": " one",
                    "probability": 0.80224609375
                },
                {
                    "start": 844.4,
                    "end": 844.72,
                    "word": " example",
                    "probability": 0.798828125
                },
                {
                    "start": 844.72,
                    "end": 844.96,
                    "word": " of",
                    "probability": 0.76416015625
                },
                {
                    "start": 844.96,
                    "end": 845.08,
                    "word": ",",
                    "probability": 0.178955078125
                },
                {
                    "start": 845.08,
                    "end": 845.42,
                    "word": " sort",
                    "probability": 0.6943359375
                },
                {
                    "start": 845.42,
                    "end": 845.5,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 845.5,
                    "end": 845.74,
                    "word": " impressive",
                    "probability": 0.320068359375
                },
                {
                    "start": 845.74,
                    "end": 846.16,
                    "word": " example",
                    "probability": 0.908203125
                },
                {
                    "start": 846.16,
                    "end": 846.22,
                    "word": ",",
                    "probability": 0.37646484375
                }
            ]
        },
        {
            "id": 139,
            "seek": 83760,
            "start": 846.22,
            "end": 849.96,
            "text": " of what transformers can do, that beg the question of whether or not they could be",
            "tokens": [
                51408,
                295,
                437,
                4088,
                433,
                393,
                360,
                11,
                300,
                4612,
                264,
                1168,
                295,
                1968,
                420,
                406,
                436,
                727,
                312,
                51600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26874999940395355,
            "compression_ratio": 1.6753731343283582,
            "no_speech_prob": 0.00628662109375,
            "words": [
                {
                    "start": 846.22,
                    "end": 846.32,
                    "word": " of",
                    "probability": 0.8984375
                },
                {
                    "start": 846.32,
                    "end": 846.42,
                    "word": " what",
                    "probability": 0.8798828125
                },
                {
                    "start": 846.42,
                    "end": 846.88,
                    "word": " transformers",
                    "probability": 0.890869140625
                },
                {
                    "start": 846.88,
                    "end": 847.02,
                    "word": " can",
                    "probability": 0.8955078125
                },
                {
                    "start": 847.02,
                    "end": 847.26,
                    "word": " do",
                    "probability": 0.91162109375
                },
                {
                    "start": 847.26,
                    "end": 848.04,
                    "word": ",",
                    "probability": 0.149169921875
                },
                {
                    "start": 848.04,
                    "end": 848.38,
                    "word": " that",
                    "probability": 0.76171875
                },
                {
                    "start": 848.38,
                    "end": 848.52,
                    "word": " beg",
                    "probability": 0.77978515625
                },
                {
                    "start": 848.52,
                    "end": 848.62,
                    "word": " the",
                    "probability": 0.64599609375
                },
                {
                    "start": 848.62,
                    "end": 848.9,
                    "word": " question",
                    "probability": 0.80712890625
                },
                {
                    "start": 848.9,
                    "end": 849.02,
                    "word": " of",
                    "probability": 0.89404296875
                },
                {
                    "start": 849.02,
                    "end": 849.12,
                    "word": " whether",
                    "probability": 0.84912109375
                },
                {
                    "start": 849.12,
                    "end": 849.28,
                    "word": " or",
                    "probability": 0.88037109375
                },
                {
                    "start": 849.28,
                    "end": 849.32,
                    "word": " not",
                    "probability": 0.8935546875
                },
                {
                    "start": 849.32,
                    "end": 849.44,
                    "word": " they",
                    "probability": 0.71630859375
                },
                {
                    "start": 849.44,
                    "end": 849.56,
                    "word": " could",
                    "probability": 0.7998046875
                },
                {
                    "start": 849.56,
                    "end": 849.96,
                    "word": " be",
                    "probability": 0.70654296875
                }
            ]
        },
        {
            "id": 140,
            "seek": 83760,
            "start": 849.96,
            "end": 854.58,
            "text": " applied usefully to phishing. And we'll show results from our experiments later in Younghoo's",
            "tokens": [
                51600,
                6456,
                764,
                2277,
                281,
                903,
                3807,
                13,
                400,
                321,
                603,
                855,
                3542,
                490,
                527,
                12050,
                1780,
                294,
                8160,
                19069,
                311,
                51832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26874999940395355,
            "compression_ratio": 1.6753731343283582,
            "no_speech_prob": 0.00628662109375,
            "words": [
                {
                    "start": 849.96,
                    "end": 850.26,
                    "word": " applied",
                    "probability": 0.7099609375
                },
                {
                    "start": 850.26,
                    "end": 850.64,
                    "word": " usefully",
                    "probability": 0.699462890625
                },
                {
                    "start": 850.64,
                    "end": 850.8,
                    "word": " to",
                    "probability": 0.89892578125
                },
                {
                    "start": 850.8,
                    "end": 851.08,
                    "word": " phishing",
                    "probability": 0.889892578125
                },
                {
                    "start": 851.08,
                    "end": 851.58,
                    "word": ".",
                    "probability": 0.43310546875
                },
                {
                    "start": 851.58,
                    "end": 851.76,
                    "word": " And",
                    "probability": 0.61767578125
                },
                {
                    "start": 851.76,
                    "end": 851.88,
                    "word": " we'll",
                    "probability": 0.892333984375
                },
                {
                    "start": 851.88,
                    "end": 852.02,
                    "word": " show",
                    "probability": 0.8291015625
                },
                {
                    "start": 852.02,
                    "end": 852.38,
                    "word": " results",
                    "probability": 0.6669921875
                },
                {
                    "start": 852.38,
                    "end": 852.56,
                    "word": " from",
                    "probability": 0.8193359375
                },
                {
                    "start": 852.56,
                    "end": 852.7,
                    "word": " our",
                    "probability": 0.81396484375
                },
                {
                    "start": 852.7,
                    "end": 853.04,
                    "word": " experiments",
                    "probability": 0.87548828125
                },
                {
                    "start": 853.04,
                    "end": 853.76,
                    "word": " later",
                    "probability": 0.873046875
                },
                {
                    "start": 853.76,
                    "end": 854.1,
                    "word": " in",
                    "probability": 0.7724609375
                },
                {
                    "start": 854.1,
                    "end": 854.58,
                    "word": " Younghoo's",
                    "probability": 0.3705647786458333
                }
            ]
        },
        {
            "id": 141,
            "seek": 86522,
            "start": 854.58,
            "end": 863.84,
            "text": " section. Here's another example of just how sort of marvelous transformer models are with respect to",
            "tokens": [
                50364,
                3541,
                13,
                1692,
                311,
                1071,
                1365,
                295,
                445,
                577,
                1333,
                295,
                34920,
                31782,
                5245,
                366,
                365,
                3104,
                281,
                50808
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23164062332361937,
            "compression_ratio": 1.5701754385964912,
            "no_speech_prob": 3.88026237487793e-05,
            "words": [
                {
                    "start": 854.58,
                    "end": 854.92,
                    "word": " section",
                    "probability": 0.5107421875
                },
                {
                    "start": 854.92,
                    "end": 855.38,
                    "word": ".",
                    "probability": 0.71630859375
                },
                {
                    "start": 855.38,
                    "end": 855.84,
                    "word": " Here's",
                    "probability": 0.9033203125
                },
                {
                    "start": 855.84,
                    "end": 856.02,
                    "word": " another",
                    "probability": 0.8232421875
                },
                {
                    "start": 856.02,
                    "end": 856.44,
                    "word": " example",
                    "probability": 0.92236328125
                },
                {
                    "start": 856.44,
                    "end": 857.04,
                    "word": " of",
                    "probability": 0.89990234375
                },
                {
                    "start": 857.04,
                    "end": 858.04,
                    "word": " just",
                    "probability": 0.87060546875
                },
                {
                    "start": 858.04,
                    "end": 858.5,
                    "word": " how",
                    "probability": 0.89892578125
                },
                {
                    "start": 858.5,
                    "end": 860.72,
                    "word": " sort",
                    "probability": 0.41845703125
                },
                {
                    "start": 860.72,
                    "end": 860.82,
                    "word": " of",
                    "probability": 0.8994140625
                },
                {
                    "start": 860.82,
                    "end": 861.2,
                    "word": " marvelous",
                    "probability": 0.6826171875
                },
                {
                    "start": 861.2,
                    "end": 861.86,
                    "word": " transformer",
                    "probability": 0.7353515625
                },
                {
                    "start": 861.86,
                    "end": 862.22,
                    "word": " models",
                    "probability": 0.80029296875
                },
                {
                    "start": 862.22,
                    "end": 862.52,
                    "word": " are",
                    "probability": 0.89306640625
                },
                {
                    "start": 862.52,
                    "end": 862.98,
                    "word": " with",
                    "probability": 0.68701171875
                },
                {
                    "start": 862.98,
                    "end": 863.28,
                    "word": " respect",
                    "probability": 0.74365234375
                },
                {
                    "start": 863.28,
                    "end": 863.84,
                    "word": " to",
                    "probability": 0.90576171875
                }
            ]
        },
        {
            "id": 142,
            "seek": 86522,
            "start": 866.38,
            "end": 872.36,
            "text": " how much more sophisticated they seem to be. So here a researcher typed into this box",
            "tokens": [
                50972,
                577,
                709,
                544,
                16950,
                436,
                1643,
                281,
                312,
                13,
                407,
                510,
                257,
                21751,
                33941,
                666,
                341,
                2424,
                51276
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23164062332361937,
            "compression_ratio": 1.5701754385964912,
            "no_speech_prob": 3.88026237487793e-05,
            "words": [
                {
                    "start": 866.38,
                    "end": 866.98,
                    "word": " how",
                    "probability": 0.8251953125
                },
                {
                    "start": 866.98,
                    "end": 867.14,
                    "word": " much",
                    "probability": 0.83642578125
                },
                {
                    "start": 867.14,
                    "end": 867.32,
                    "word": " more",
                    "probability": 0.87060546875
                },
                {
                    "start": 867.32,
                    "end": 867.72,
                    "word": " sophisticated",
                    "probability": 0.7578125
                },
                {
                    "start": 867.72,
                    "end": 867.94,
                    "word": " they",
                    "probability": 0.73974609375
                },
                {
                    "start": 867.94,
                    "end": 868.06,
                    "word": " seem",
                    "probability": 0.83984375
                },
                {
                    "start": 868.06,
                    "end": 868.18,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 868.18,
                    "end": 868.32,
                    "word": " be",
                    "probability": 0.90625
                },
                {
                    "start": 868.32,
                    "end": 868.46,
                    "word": ".",
                    "probability": 0.9111328125
                },
                {
                    "start": 868.46,
                    "end": 868.46,
                    "word": " So",
                    "probability": 0.82958984375
                },
                {
                    "start": 868.46,
                    "end": 868.86,
                    "word": " here",
                    "probability": 0.73046875
                },
                {
                    "start": 870.16,
                    "end": 870.5,
                    "word": " a researcher",
                    "probability": 0.8203125
                },
                {
                    "start": 870.5,
                    "end": 870.84,
                    "word": " typed",
                    "probability": 0.75927734375
                },
                {
                    "start": 870.84,
                    "end": 870.98,
                    "word": " into",
                    "probability": 0.61572265625
                },
                {
                    "start": 870.98,
                    "end": 872.0,
                    "word": " this",
                    "probability": 0.86474609375
                },
                {
                    "start": 872.0,
                    "end": 872.36,
                    "word": " box",
                    "probability": 0.80126953125
                }
            ]
        },
        {
            "id": 143,
            "seek": 86522,
            "start": 873.1,
            "end": 876.66,
            "text": " a description of some HTML that he would like the transformer models to generate.",
            "tokens": [
                51276,
                257,
                3855,
                295,
                512,
                17995,
                300,
                415,
                576,
                411,
                264,
                31782,
                5245,
                281,
                8460,
                13,
                51468
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23164062332361937,
            "compression_ratio": 1.5701754385964912,
            "no_speech_prob": 3.88026237487793e-05,
            "words": [
                {
                    "start": 873.1,
                    "end": 873.46,
                    "word": " a description",
                    "probability": 0.7919921875
                },
                {
                    "start": 873.46,
                    "end": 873.66,
                    "word": " of",
                    "probability": 0.89990234375
                },
                {
                    "start": 873.66,
                    "end": 873.8,
                    "word": " some",
                    "probability": 0.802734375
                },
                {
                    "start": 873.8,
                    "end": 874.12,
                    "word": " HTML",
                    "probability": 0.830078125
                },
                {
                    "start": 874.12,
                    "end": 874.36,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 874.36,
                    "end": 874.46,
                    "word": " he",
                    "probability": 0.90625
                },
                {
                    "start": 874.46,
                    "end": 874.6,
                    "word": " would",
                    "probability": 0.8427734375
                },
                {
                    "start": 874.6,
                    "end": 874.88,
                    "word": " like",
                    "probability": 0.8828125
                },
                {
                    "start": 874.88,
                    "end": 875.06,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 875.06,
                    "end": 875.42,
                    "word": " transformer",
                    "probability": 0.8359375
                },
                {
                    "start": 875.42,
                    "end": 875.76,
                    "word": " models",
                    "probability": 0.437255859375
                },
                {
                    "start": 875.76,
                    "end": 875.88,
                    "word": " to",
                    "probability": 0.84130859375
                },
                {
                    "start": 875.88,
                    "end": 876.26,
                    "word": " generate",
                    "probability": 0.8447265625
                },
                {
                    "start": 876.26,
                    "end": 876.66,
                    "word": ".",
                    "probability": 0.9365234375
                }
            ]
        },
        {
            "id": 144,
            "seek": 86522,
            "start": 876.66,
            "end": 882.2,
            "text": " He writes, a button with the color of Donald Trump's hair. And the model actually writes,",
            "tokens": [
                51468,
                634,
                13657,
                11,
                257,
                2960,
                365,
                264,
                2017,
                295,
                8632,
                3899,
                311,
                2578,
                13,
                400,
                264,
                2316,
                767,
                13657,
                11,
                51724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23164062332361937,
            "compression_ratio": 1.5701754385964912,
            "no_speech_prob": 3.88026237487793e-05,
            "words": [
                {
                    "start": 876.66,
                    "end": 876.94,
                    "word": " He",
                    "probability": 0.88525390625
                },
                {
                    "start": 876.94,
                    "end": 877.24,
                    "word": " writes",
                    "probability": 0.81640625
                },
                {
                    "start": 877.24,
                    "end": 877.44,
                    "word": ",",
                    "probability": 0.498291015625
                },
                {
                    "start": 877.78,
                    "end": 878.0,
                    "word": " a button",
                    "probability": 0.8203125
                },
                {
                    "start": 878.0,
                    "end": 878.26,
                    "word": " with",
                    "probability": 0.8134765625
                },
                {
                    "start": 878.26,
                    "end": 878.38,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 878.38,
                    "end": 878.58,
                    "word": " color",
                    "probability": 0.8095703125
                },
                {
                    "start": 878.58,
                    "end": 878.74,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 878.74,
                    "end": 878.94,
                    "word": " Donald",
                    "probability": 0.8173828125
                },
                {
                    "start": 878.94,
                    "end": 879.32,
                    "word": " Trump's",
                    "probability": 0.935791015625
                },
                {
                    "start": 879.32,
                    "end": 879.54,
                    "word": " hair",
                    "probability": 0.865234375
                },
                {
                    "start": 879.54,
                    "end": 880.14,
                    "word": ".",
                    "probability": 0.85693359375
                },
                {
                    "start": 880.64,
                    "end": 880.82,
                    "word": " And",
                    "probability": 0.837890625
                },
                {
                    "start": 880.82,
                    "end": 880.92,
                    "word": " the",
                    "probability": 0.6943359375
                },
                {
                    "start": 880.92,
                    "end": 881.06,
                    "word": " model",
                    "probability": 0.837890625
                },
                {
                    "start": 881.06,
                    "end": 881.34,
                    "word": " actually",
                    "probability": 0.78515625
                },
                {
                    "start": 881.34,
                    "end": 881.86,
                    "word": " writes",
                    "probability": 0.78759765625
                },
                {
                    "start": 881.86,
                    "end": 882.2,
                    "word": ",",
                    "probability": 0.31591796875
                }
            ]
        },
        {
            "id": 145,
            "seek": 89064,
            "start": 882.82,
            "end": 889.04,
            "text": " as a text completion, some HTML and CSS. It seems to understand that Donald Trump has yellow hair,",
            "tokens": [
                50408,
                382,
                257,
                2487,
                19372,
                11,
                512,
                17995,
                293,
                24387,
                13,
                467,
                2544,
                281,
                1223,
                300,
                8632,
                3899,
                575,
                5566,
                2578,
                11,
                50684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23355263691199454,
            "compression_ratio": 1.5224489795918368,
            "no_speech_prob": 0.0008420944213867188,
            "words": [
                {
                    "start": 882.82,
                    "end": 883.26,
                    "word": " as",
                    "probability": 0.40087890625
                },
                {
                    "start": 883.38,
                    "end": 883.56,
                    "word": " a text",
                    "probability": 0.8427734375
                },
                {
                    "start": 883.56,
                    "end": 884.06,
                    "word": " completion",
                    "probability": 0.71728515625
                },
                {
                    "start": 884.06,
                    "end": 884.66,
                    "word": ",",
                    "probability": 0.59033203125
                },
                {
                    "start": 884.66,
                    "end": 884.9,
                    "word": " some",
                    "probability": 0.7392578125
                },
                {
                    "start": 884.9,
                    "end": 885.26,
                    "word": " HTML",
                    "probability": 0.85205078125
                },
                {
                    "start": 885.26,
                    "end": 885.44,
                    "word": " and",
                    "probability": 0.88427734375
                },
                {
                    "start": 885.44,
                    "end": 885.7,
                    "word": " CSS",
                    "probability": 0.84912109375
                },
                {
                    "start": 885.7,
                    "end": 886.12,
                    "word": ".",
                    "probability": 0.4931640625
                },
                {
                    "start": 886.22,
                    "end": 886.66,
                    "word": " It",
                    "probability": 0.90234375
                },
                {
                    "start": 886.66,
                    "end": 886.84,
                    "word": " seems",
                    "probability": 0.7265625
                },
                {
                    "start": 886.84,
                    "end": 886.98,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 886.98,
                    "end": 887.34,
                    "word": " understand",
                    "probability": 0.736328125
                },
                {
                    "start": 887.34,
                    "end": 887.48,
                    "word": " that",
                    "probability": 0.84033203125
                },
                {
                    "start": 887.48,
                    "end": 887.68,
                    "word": " Donald",
                    "probability": 0.85595703125
                },
                {
                    "start": 887.68,
                    "end": 887.9,
                    "word": " Trump",
                    "probability": 0.865234375
                },
                {
                    "start": 887.9,
                    "end": 888.08,
                    "word": " has",
                    "probability": 0.8896484375
                },
                {
                    "start": 888.08,
                    "end": 888.32,
                    "word": " yellow",
                    "probability": 0.8125
                },
                {
                    "start": 888.32,
                    "end": 888.58,
                    "word": " hair",
                    "probability": 0.89453125
                },
                {
                    "start": 888.58,
                    "end": 889.04,
                    "word": ",",
                    "probability": 0.272216796875
                }
            ]
        },
        {
            "id": 146,
            "seek": 89064,
            "start": 889.04,
            "end": 897.0,
            "text": " and it makes a button, and it writes valid HTML and CSS. Which is, I think, just",
            "tokens": [
                50716,
                293,
                309,
                1669,
                257,
                2960,
                11,
                293,
                309,
                13657,
                7363,
                17995,
                293,
                24387,
                13,
                3013,
                307,
                11,
                286,
                519,
                11,
                445,
                51128
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23355263691199454,
            "compression_ratio": 1.5224489795918368,
            "no_speech_prob": 0.0008420944213867188,
            "words": [
                {
                    "start": 889.04,
                    "end": 889.74,
                    "word": " and",
                    "probability": 0.8896484375
                },
                {
                    "start": 889.74,
                    "end": 890.52,
                    "word": " it",
                    "probability": 0.8251953125
                },
                {
                    "start": 890.52,
                    "end": 890.7,
                    "word": " makes",
                    "probability": 0.72607421875
                },
                {
                    "start": 890.82,
                    "end": 891.02,
                    "word": " a button",
                    "probability": 0.8232421875
                },
                {
                    "start": 891.02,
                    "end": 891.16,
                    "word": ",",
                    "probability": 0.320556640625
                },
                {
                    "start": 891.16,
                    "end": 891.18,
                    "word": " and",
                    "probability": 0.88525390625
                },
                {
                    "start": 891.18,
                    "end": 891.28,
                    "word": " it",
                    "probability": 0.87841796875
                },
                {
                    "start": 891.28,
                    "end": 891.46,
                    "word": " writes",
                    "probability": 0.75537109375
                },
                {
                    "start": 891.46,
                    "end": 891.82,
                    "word": " valid",
                    "probability": 0.5927734375
                },
                {
                    "start": 891.82,
                    "end": 892.22,
                    "word": " HTML",
                    "probability": 0.93115234375
                },
                {
                    "start": 892.22,
                    "end": 892.88,
                    "word": " and",
                    "probability": 0.87109375
                },
                {
                    "start": 892.88,
                    "end": 893.54,
                    "word": " CSS",
                    "probability": 0.796875
                },
                {
                    "start": 893.54,
                    "end": 893.98,
                    "word": ".",
                    "probability": 0.273681640625
                },
                {
                    "start": 894.78,
                    "end": 895.22,
                    "word": " Which",
                    "probability": 0.81494140625
                },
                {
                    "start": 895.22,
                    "end": 895.52,
                    "word": " is",
                    "probability": 0.89453125
                },
                {
                    "start": 895.52,
                    "end": 895.64,
                    "word": ",",
                    "probability": 0.6787109375
                },
                {
                    "start": 895.64,
                    "end": 896.68,
                    "word": " I",
                    "probability": 0.9892578125
                },
                {
                    "start": 896.68,
                    "end": 896.8,
                    "word": " think",
                    "probability": 0.80712890625
                },
                {
                    "start": 896.8,
                    "end": 896.9,
                    "word": ",",
                    "probability": 0.66650390625
                },
                {
                    "start": 896.9,
                    "end": 897.0,
                    "word": " just",
                    "probability": 0.8623046875
                }
            ]
        },
        {
            "id": 147,
            "seek": 89064,
            "start": 898.72,
            "end": 902.82,
            "text": " impressive in the absolute sense, but also for folks who've been in the NLP space for a while,",
            "tokens": [
                51192,
                8992,
                294,
                264,
                8236,
                2020,
                11,
                457,
                611,
                337,
                4024,
                567,
                600,
                668,
                294,
                264,
                426,
                45196,
                1901,
                337,
                257,
                1339,
                11,
                51388
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23355263691199454,
            "compression_ratio": 1.5224489795918368,
            "no_speech_prob": 0.0008420944213867188,
            "words": [
                {
                    "start": 898.72,
                    "end": 899.16,
                    "word": " impressive",
                    "probability": 0.52099609375
                },
                {
                    "start": 899.16,
                    "end": 899.4,
                    "word": " in",
                    "probability": 0.8955078125
                },
                {
                    "start": 899.4,
                    "end": 899.54,
                    "word": " the",
                    "probability": 0.78955078125
                },
                {
                    "start": 899.54,
                    "end": 899.76,
                    "word": " absolute",
                    "probability": 0.77099609375
                },
                {
                    "start": 899.76,
                    "end": 900.1,
                    "word": " sense",
                    "probability": 0.74462890625
                },
                {
                    "start": 900.1,
                    "end": 900.18,
                    "word": ",",
                    "probability": 0.8349609375
                },
                {
                    "start": 900.18,
                    "end": 900.26,
                    "word": " but",
                    "probability": 0.8994140625
                },
                {
                    "start": 900.26,
                    "end": 900.54,
                    "word": " also",
                    "probability": 0.81103515625
                },
                {
                    "start": 900.54,
                    "end": 901.02,
                    "word": " for",
                    "probability": 0.6572265625
                },
                {
                    "start": 901.02,
                    "end": 901.3,
                    "word": " folks",
                    "probability": 0.85888671875
                },
                {
                    "start": 901.3,
                    "end": 901.44,
                    "word": " who've",
                    "probability": 0.75244140625
                },
                {
                    "start": 901.44,
                    "end": 901.54,
                    "word": " been",
                    "probability": 0.8154296875
                },
                {
                    "start": 901.54,
                    "end": 901.64,
                    "word": " in",
                    "probability": 0.8916015625
                },
                {
                    "start": 901.64,
                    "end": 901.72,
                    "word": " the",
                    "probability": 0.7841796875
                },
                {
                    "start": 901.72,
                    "end": 901.96,
                    "word": " NLP",
                    "probability": 0.8955078125
                },
                {
                    "start": 901.96,
                    "end": 902.34,
                    "word": " space",
                    "probability": 0.83251953125
                },
                {
                    "start": 902.34,
                    "end": 902.5,
                    "word": " for",
                    "probability": 0.88623046875
                },
                {
                    "start": 902.58,
                    "end": 902.74,
                    "word": " a while",
                    "probability": 0.90576171875
                },
                {
                    "start": 902.74,
                    "end": 902.82,
                    "word": ",",
                    "probability": 0.85009765625
                }
            ]
        },
        {
            "id": 148,
            "seek": 89064,
            "start": 902.82,
            "end": 907.62,
            "text": " it's impressive relatively. Five, ten years ago, if you would have shown somebody that we could do",
            "tokens": [
                51388,
                309,
                311,
                8992,
                7226,
                13,
                9436,
                11,
                2064,
                924,
                2057,
                11,
                498,
                291,
                576,
                362,
                4898,
                2618,
                300,
                321,
                727,
                360,
                51632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23355263691199454,
            "compression_ratio": 1.5224489795918368,
            "no_speech_prob": 0.0008420944213867188,
            "words": [
                {
                    "start": 902.82,
                    "end": 902.94,
                    "word": " it's",
                    "probability": 0.947021484375
                },
                {
                    "start": 902.94,
                    "end": 903.34,
                    "word": " impressive",
                    "probability": 0.67626953125
                },
                {
                    "start": 903.34,
                    "end": 903.92,
                    "word": " relatively",
                    "probability": 0.6181640625
                },
                {
                    "start": 903.92,
                    "end": 904.26,
                    "word": ".",
                    "probability": 0.72119140625
                },
                {
                    "start": 904.26,
                    "end": 904.64,
                    "word": " Five",
                    "probability": 0.74072265625
                },
                {
                    "start": 904.64,
                    "end": 904.76,
                    "word": ",",
                    "probability": 0.96044921875
                },
                {
                    "start": 904.76,
                    "end": 904.86,
                    "word": " ten",
                    "probability": 0.357177734375
                },
                {
                    "start": 904.86,
                    "end": 905.06,
                    "word": " years",
                    "probability": 0.8818359375
                },
                {
                    "start": 905.06,
                    "end": 905.42,
                    "word": " ago",
                    "probability": 0.80859375
                },
                {
                    "start": 905.42,
                    "end": 905.82,
                    "word": ",",
                    "probability": 0.88134765625
                },
                {
                    "start": 905.82,
                    "end": 906.12,
                    "word": " if",
                    "probability": 0.85693359375
                },
                {
                    "start": 906.12,
                    "end": 906.16,
                    "word": " you",
                    "probability": 0.88671875
                },
                {
                    "start": 906.16,
                    "end": 906.26,
                    "word": " would",
                    "probability": 0.8076171875
                },
                {
                    "start": 906.26,
                    "end": 906.32,
                    "word": " have",
                    "probability": 0.62353515625
                },
                {
                    "start": 906.32,
                    "end": 906.48,
                    "word": " shown",
                    "probability": 0.89990234375
                },
                {
                    "start": 906.48,
                    "end": 906.8,
                    "word": " somebody",
                    "probability": 0.78564453125
                },
                {
                    "start": 906.8,
                    "end": 907.06,
                    "word": " that",
                    "probability": 0.87939453125
                },
                {
                    "start": 907.06,
                    "end": 907.34,
                    "word": " we",
                    "probability": 0.8896484375
                },
                {
                    "start": 907.34,
                    "end": 907.48,
                    "word": " could",
                    "probability": 0.80908203125
                },
                {
                    "start": 907.48,
                    "end": 907.62,
                    "word": " do",
                    "probability": 0.908203125
                }
            ]
        },
        {
            "id": 149,
            "seek": 91214,
            "start": 907.62,
            "end": 913.82,
            "text": " this in 2020, I think they would have been a bit incredulous. This is a big step for you.",
            "tokens": [
                50364,
                341,
                294,
                4808,
                11,
                286,
                519,
                436,
                576,
                362,
                668,
                257,
                857,
                3267,
                6893,
                13,
                639,
                307,
                257,
                955,
                1823,
                337,
                291,
                13,
                50648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2271634585909791,
            "compression_ratio": 1.5129310344827587,
            "no_speech_prob": 0.0014095306396484375,
            "words": [
                {
                    "start": 907.62,
                    "end": 907.9,
                    "word": " this",
                    "probability": 0.398193359375
                },
                {
                    "start": 907.9,
                    "end": 908.52,
                    "word": " in",
                    "probability": 0.7607421875
                },
                {
                    "start": 908.52,
                    "end": 908.96,
                    "word": " 2020",
                    "probability": 0.826171875
                },
                {
                    "start": 908.96,
                    "end": 909.22,
                    "word": ",",
                    "probability": 0.6005859375
                },
                {
                    "start": 909.22,
                    "end": 909.3,
                    "word": " I",
                    "probability": 0.93798828125
                },
                {
                    "start": 909.3,
                    "end": 909.46,
                    "word": " think",
                    "probability": 0.8115234375
                },
                {
                    "start": 909.46,
                    "end": 909.6,
                    "word": " they",
                    "probability": 0.72216796875
                },
                {
                    "start": 909.6,
                    "end": 909.72,
                    "word": " would",
                    "probability": 0.85107421875
                },
                {
                    "start": 909.72,
                    "end": 909.9,
                    "word": " have",
                    "probability": 0.82861328125
                },
                {
                    "start": 909.9,
                    "end": 911.3,
                    "word": " been",
                    "probability": 0.355712890625
                },
                {
                    "start": 911.44,
                    "end": 911.54,
                    "word": " a bit",
                    "probability": 0.88525390625
                },
                {
                    "start": 911.54,
                    "end": 912.14,
                    "word": " incredulous",
                    "probability": 0.88330078125
                },
                {
                    "start": 912.14,
                    "end": 912.62,
                    "word": ".",
                    "probability": 0.7861328125
                },
                {
                    "start": 912.62,
                    "end": 912.62,
                    "word": " This",
                    "probability": 0.287841796875
                },
                {
                    "start": 912.62,
                    "end": 912.68,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 912.76,
                    "end": 912.88,
                    "word": " a big",
                    "probability": 0.892578125
                },
                {
                    "start": 912.88,
                    "end": 913.12,
                    "word": " step",
                    "probability": 0.869140625
                },
                {
                    "start": 913.12,
                    "end": 913.24,
                    "word": " for",
                    "probability": 0.382080078125
                },
                {
                    "start": 913.24,
                    "end": 913.36,
                    "word": " you",
                    "probability": 0.39111328125
                },
                {
                    "start": 913.36,
                    "end": 913.82,
                    "word": ".",
                    "probability": 0.87255859375
                }
            ]
        },
        {
            "id": 150,
            "seek": 91214,
            "start": 913.96,
            "end": 918.28,
            "text": " Applications like these represent a big step forward in natural language processing. And",
            "tokens": [
                50680,
                26519,
                763,
                411,
                613,
                2906,
                257,
                955,
                1823,
                2128,
                294,
                3303,
                2856,
                9007,
                13,
                400,
                50900
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2271634585909791,
            "compression_ratio": 1.5129310344827587,
            "no_speech_prob": 0.0014095306396484375,
            "words": [
                {
                    "start": 913.96,
                    "end": 914.44,
                    "word": " Applications",
                    "probability": 0.888916015625
                },
                {
                    "start": 914.44,
                    "end": 914.64,
                    "word": " like",
                    "probability": 0.88427734375
                },
                {
                    "start": 914.64,
                    "end": 914.98,
                    "word": " these",
                    "probability": 0.681640625
                },
                {
                    "start": 914.98,
                    "end": 916.3,
                    "word": " represent",
                    "probability": 0.78857421875
                },
                {
                    "start": 916.42,
                    "end": 916.54,
                    "word": " a big",
                    "probability": 0.892578125
                },
                {
                    "start": 916.54,
                    "end": 916.72,
                    "word": " step",
                    "probability": 0.80029296875
                },
                {
                    "start": 916.72,
                    "end": 917.0,
                    "word": " forward",
                    "probability": 0.82421875
                },
                {
                    "start": 917.0,
                    "end": 917.14,
                    "word": " in",
                    "probability": 0.89599609375
                },
                {
                    "start": 917.14,
                    "end": 917.34,
                    "word": " natural",
                    "probability": 0.79638671875
                },
                {
                    "start": 917.34,
                    "end": 917.58,
                    "word": " language",
                    "probability": 0.66796875
                },
                {
                    "start": 917.58,
                    "end": 917.98,
                    "word": " processing",
                    "probability": 0.82470703125
                },
                {
                    "start": 917.98,
                    "end": 918.22,
                    "word": ".",
                    "probability": 0.57568359375
                },
                {
                    "start": 918.22,
                    "end": 918.28,
                    "word": " And",
                    "probability": 0.740234375
                }
            ]
        },
        {
            "id": 151,
            "seek": 91214,
            "start": 918.28,
            "end": 924.64,
            "text": " they're due, again, to this new idea called a transformer block. That's the key building",
            "tokens": [
                50900,
                436,
                434,
                3462,
                11,
                797,
                11,
                281,
                341,
                777,
                1558,
                1219,
                257,
                31782,
                3461,
                13,
                663,
                311,
                264,
                2141,
                2390,
                51212
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2271634585909791,
            "compression_ratio": 1.5129310344827587,
            "no_speech_prob": 0.0014095306396484375,
            "words": [
                {
                    "start": 918.28,
                    "end": 918.6,
                    "word": " they're",
                    "probability": 0.81103515625
                },
                {
                    "start": 918.6,
                    "end": 918.72,
                    "word": " due",
                    "probability": 0.8857421875
                },
                {
                    "start": 918.72,
                    "end": 918.98,
                    "word": ",",
                    "probability": 0.36572265625
                },
                {
                    "start": 918.98,
                    "end": 918.98,
                    "word": " again",
                    "probability": 0.921875
                },
                {
                    "start": 918.98,
                    "end": 919.04,
                    "word": ",",
                    "probability": 0.986328125
                },
                {
                    "start": 919.04,
                    "end": 919.12,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 919.12,
                    "end": 919.3,
                    "word": " this",
                    "probability": 0.88330078125
                },
                {
                    "start": 919.3,
                    "end": 919.44,
                    "word": " new",
                    "probability": 0.8134765625
                },
                {
                    "start": 919.44,
                    "end": 919.78,
                    "word": " idea",
                    "probability": 0.7734375
                },
                {
                    "start": 919.78,
                    "end": 920.72,
                    "word": " called",
                    "probability": 0.76171875
                },
                {
                    "start": 921.72,
                    "end": 922.04,
                    "word": " a transformer",
                    "probability": 0.7138671875
                },
                {
                    "start": 922.04,
                    "end": 922.5,
                    "word": " block",
                    "probability": 0.80908203125
                },
                {
                    "start": 922.5,
                    "end": 922.98,
                    "word": ".",
                    "probability": 0.94091796875
                },
                {
                    "start": 923.2,
                    "end": 923.56,
                    "word": " That's",
                    "probability": 0.886962890625
                },
                {
                    "start": 923.56,
                    "end": 924.18,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 924.18,
                    "end": 924.34,
                    "word": " key",
                    "probability": 0.7890625
                },
                {
                    "start": 924.34,
                    "end": 924.64,
                    "word": " building",
                    "probability": 0.79248046875
                }
            ]
        },
        {
            "id": 152,
            "seek": 91214,
            "start": 924.64,
            "end": 929.12,
            "text": " block out of which models like this model, which is called GPT-3. It's from OpenAI,",
            "tokens": [
                51212,
                3461,
                484,
                295,
                597,
                5245,
                411,
                341,
                2316,
                11,
                597,
                307,
                1219,
                26039,
                51,
                12,
                18,
                13,
                467,
                311,
                490,
                7238,
                48698,
                11,
                51432
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2271634585909791,
            "compression_ratio": 1.5129310344827587,
            "no_speech_prob": 0.0014095306396484375,
            "words": [
                {
                    "start": 924.64,
                    "end": 924.9,
                    "word": " block",
                    "probability": 0.82470703125
                },
                {
                    "start": 924.9,
                    "end": 925.02,
                    "word": " out",
                    "probability": 0.76708984375
                },
                {
                    "start": 925.02,
                    "end": 925.1,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 925.1,
                    "end": 925.34,
                    "word": " which",
                    "probability": 0.91796875
                },
                {
                    "start": 925.34,
                    "end": 925.96,
                    "word": " models",
                    "probability": 0.81591796875
                },
                {
                    "start": 925.96,
                    "end": 926.2,
                    "word": " like",
                    "probability": 0.732421875
                },
                {
                    "start": 926.2,
                    "end": 926.54,
                    "word": " this",
                    "probability": 0.87841796875
                },
                {
                    "start": 926.54,
                    "end": 926.72,
                    "word": " model",
                    "probability": 0.83984375
                },
                {
                    "start": 926.72,
                    "end": 926.78,
                    "word": ",",
                    "probability": 0.8525390625
                },
                {
                    "start": 926.78,
                    "end": 926.88,
                    "word": " which",
                    "probability": 0.9169921875
                },
                {
                    "start": 926.88,
                    "end": 926.98,
                    "word": " is",
                    "probability": 0.90283203125
                },
                {
                    "start": 926.98,
                    "end": 927.14,
                    "word": " called",
                    "probability": 0.8271484375
                },
                {
                    "start": 927.14,
                    "end": 927.54,
                    "word": " GPT",
                    "probability": 0.8974609375
                },
                {
                    "start": 927.54,
                    "end": 927.8,
                    "word": "-3",
                    "probability": 0.91357421875
                },
                {
                    "start": 927.8,
                    "end": 927.88,
                    "word": ".",
                    "probability": 0.42236328125
                },
                {
                    "start": 927.88,
                    "end": 928.12,
                    "word": " It's",
                    "probability": 0.9453125
                },
                {
                    "start": 928.12,
                    "end": 928.32,
                    "word": " from",
                    "probability": 0.81396484375
                },
                {
                    "start": 928.32,
                    "end": 928.88,
                    "word": " OpenAI",
                    "probability": 0.78759765625
                },
                {
                    "start": 928.88,
                    "end": 929.12,
                    "word": ",",
                    "probability": 0.955078125
                }
            ]
        },
        {
            "id": 153,
            "seek": 93706,
            "start": 929.7,
            "end": 939.34,
            "text": " which is an AI research lab, are based. Okay. So, moving on.",
            "tokens": [
                50396,
                597,
                307,
                364,
                7318,
                2132,
                2715,
                11,
                366,
                2361,
                13,
                1033,
                13,
                407,
                11,
                2684,
                322,
                13,
                50836
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21310096153846153,
            "compression_ratio": 1.3823529411764706,
            "no_speech_prob": 0.0007090568542480469,
            "words": [
                {
                    "start": 929.7,
                    "end": 930.04,
                    "word": " which",
                    "probability": 0.7001953125
                },
                {
                    "start": 930.04,
                    "end": 930.16,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 930.16,
                    "end": 930.3,
                    "word": " an",
                    "probability": 0.2039794921875
                },
                {
                    "start": 930.3,
                    "end": 930.56,
                    "word": " AI",
                    "probability": 0.90234375
                },
                {
                    "start": 930.56,
                    "end": 931.78,
                    "word": " research",
                    "probability": 0.73095703125
                },
                {
                    "start": 931.78,
                    "end": 932.14,
                    "word": " lab",
                    "probability": 0.8974609375
                },
                {
                    "start": 932.14,
                    "end": 932.96,
                    "word": ",",
                    "probability": 0.274658203125
                },
                {
                    "start": 932.96,
                    "end": 933.16,
                    "word": " are",
                    "probability": 0.8515625
                },
                {
                    "start": 933.16,
                    "end": 933.68,
                    "word": " based",
                    "probability": 0.8330078125
                },
                {
                    "start": 933.68,
                    "end": 934.36,
                    "word": ".",
                    "probability": 0.9833984375
                },
                {
                    "start": 935.46,
                    "end": 935.68,
                    "word": " Okay",
                    "probability": 0.66015625
                },
                {
                    "start": 935.68,
                    "end": 936.36,
                    "word": ".",
                    "probability": 0.48046875
                },
                {
                    "start": 936.84,
                    "end": 937.52,
                    "word": " So",
                    "probability": 0.82958984375
                },
                {
                    "start": 937.52,
                    "end": 937.8,
                    "word": ",",
                    "probability": 0.333740234375
                },
                {
                    "start": 937.8,
                    "end": 938.32,
                    "word": " moving",
                    "probability": 0.86328125
                },
                {
                    "start": 938.32,
                    "end": 938.66,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 938.66,
                    "end": 939.34,
                    "word": ".",
                    "probability": 0.92138671875
                }
            ]
        },
        {
            "id": 154,
            "seek": 93706,
            "start": 939.9,
            "end": 947.18,
            "text": " So, I want to go back a little bit in the history of NLP as a way of describing how",
            "tokens": [
                50900,
                407,
                11,
                286,
                528,
                281,
                352,
                646,
                257,
                707,
                857,
                294,
                264,
                2503,
                295,
                426,
                45196,
                382,
                257,
                636,
                295,
                16141,
                577,
                51268
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21310096153846153,
            "compression_ratio": 1.3823529411764706,
            "no_speech_prob": 0.0007090568542480469,
            "words": [
                {
                    "start": 939.9,
                    "end": 940.14,
                    "word": " So",
                    "probability": 0.7587890625
                },
                {
                    "start": 940.14,
                    "end": 940.24,
                    "word": ",",
                    "probability": 0.62158203125
                },
                {
                    "start": 940.24,
                    "end": 940.58,
                    "word": " I",
                    "probability": 0.99853515625
                },
                {
                    "start": 940.58,
                    "end": 940.7,
                    "word": " want",
                    "probability": 0.424072265625
                },
                {
                    "start": 940.7,
                    "end": 940.76,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 940.76,
                    "end": 940.86,
                    "word": " go",
                    "probability": 0.896484375
                },
                {
                    "start": 940.86,
                    "end": 941.06,
                    "word": " back",
                    "probability": 0.8125
                },
                {
                    "start": 941.22,
                    "end": 941.38,
                    "word": " a little",
                    "probability": 0.8046875
                },
                {
                    "start": 941.38,
                    "end": 941.82,
                    "word": " bit",
                    "probability": 0.87890625
                },
                {
                    "start": 941.82,
                    "end": 942.64,
                    "word": " in",
                    "probability": 0.82373046875
                },
                {
                    "start": 942.64,
                    "end": 942.76,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 942.76,
                    "end": 943.0,
                    "word": " history",
                    "probability": 0.83154296875
                },
                {
                    "start": 943.0,
                    "end": 943.12,
                    "word": " of",
                    "probability": 0.8994140625
                },
                {
                    "start": 943.12,
                    "end": 943.46,
                    "word": " NLP",
                    "probability": 0.9091796875
                },
                {
                    "start": 943.46,
                    "end": 943.86,
                    "word": " as",
                    "probability": 0.5517578125
                },
                {
                    "start": 944.96,
                    "end": 945.24,
                    "word": " a way",
                    "probability": 0.90283203125
                },
                {
                    "start": 945.24,
                    "end": 945.9,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 945.9,
                    "end": 946.86,
                    "word": " describing",
                    "probability": 0.7890625
                },
                {
                    "start": 946.86,
                    "end": 947.18,
                    "word": " how",
                    "probability": 0.90087890625
                }
            ]
        },
        {
            "id": 155,
            "seek": 93706,
            "start": 947.18,
            "end": 954.04,
            "text": " transformers are new and different. So, I assume a substantial chunk of our audience today",
            "tokens": [
                51268,
                4088,
                433,
                366,
                777,
                293,
                819,
                13,
                407,
                11,
                286,
                6552,
                257,
                16726,
                16635,
                295,
                527,
                4034,
                965,
                51608
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21310096153846153,
            "compression_ratio": 1.3823529411764706,
            "no_speech_prob": 0.0007090568542480469,
            "words": [
                {
                    "start": 947.18,
                    "end": 947.8,
                    "word": " transformers",
                    "probability": 0.833251953125
                },
                {
                    "start": 947.8,
                    "end": 947.96,
                    "word": " are",
                    "probability": 0.8935546875
                },
                {
                    "start": 947.96,
                    "end": 948.34,
                    "word": " new",
                    "probability": 0.79052734375
                },
                {
                    "start": 948.34,
                    "end": 948.5,
                    "word": " and",
                    "probability": 0.8935546875
                },
                {
                    "start": 948.5,
                    "end": 948.78,
                    "word": " different",
                    "probability": 0.6923828125
                },
                {
                    "start": 948.78,
                    "end": 949.46,
                    "word": ".",
                    "probability": 0.97705078125
                },
                {
                    "start": 949.92,
                    "end": 950.12,
                    "word": " So",
                    "probability": 0.8505859375
                },
                {
                    "start": 950.12,
                    "end": 950.34,
                    "word": ",",
                    "probability": 0.88330078125
                },
                {
                    "start": 950.34,
                    "end": 950.6,
                    "word": " I",
                    "probability": 0.9912109375
                },
                {
                    "start": 950.6,
                    "end": 950.9,
                    "word": " assume",
                    "probability": 0.779296875
                },
                {
                    "start": 951.54,
                    "end": 951.88,
                    "word": " a substantial",
                    "probability": 0.77734375
                },
                {
                    "start": 951.88,
                    "end": 952.58,
                    "word": " chunk",
                    "probability": 0.81201171875
                },
                {
                    "start": 952.58,
                    "end": 952.84,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 952.84,
                    "end": 953.26,
                    "word": " our",
                    "probability": 0.80810546875
                },
                {
                    "start": 953.26,
                    "end": 953.6,
                    "word": " audience",
                    "probability": 0.79443359375
                },
                {
                    "start": 953.6,
                    "end": 954.04,
                    "word": " today",
                    "probability": 0.7939453125
                }
            ]
        },
        {
            "id": 156,
            "seek": 96260,
            "start": 954.64,
            "end": 962.72,
            "text": " has studied basic machine learning on text. And typically, in like a machine learning 101",
            "tokens": [
                50396,
                575,
                9454,
                3875,
                3479,
                2539,
                322,
                2487,
                13,
                400,
                5850,
                11,
                294,
                411,
                257,
                3479,
                2539,
                21055,
                50800
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21893601651702607,
            "compression_ratio": 1.6936936936936937,
            "no_speech_prob": 0.0007786750793457031,
            "words": [
                {
                    "start": 954.64,
                    "end": 955.08,
                    "word": " has",
                    "probability": 0.59326171875
                },
                {
                    "start": 955.08,
                    "end": 955.52,
                    "word": " studied",
                    "probability": 0.87109375
                },
                {
                    "start": 955.52,
                    "end": 956.8,
                    "word": " basic",
                    "probability": 0.92626953125
                },
                {
                    "start": 956.8,
                    "end": 958.18,
                    "word": " machine",
                    "probability": 0.701171875
                },
                {
                    "start": 958.18,
                    "end": 958.52,
                    "word": " learning",
                    "probability": 0.83935546875
                },
                {
                    "start": 958.52,
                    "end": 958.84,
                    "word": " on",
                    "probability": 0.90283203125
                },
                {
                    "start": 958.84,
                    "end": 959.32,
                    "word": " text",
                    "probability": 0.91455078125
                },
                {
                    "start": 959.32,
                    "end": 959.76,
                    "word": ".",
                    "probability": 0.49609375
                },
                {
                    "start": 960.58,
                    "end": 960.74,
                    "word": " And",
                    "probability": 0.72314453125
                },
                {
                    "start": 960.74,
                    "end": 961.42,
                    "word": " typically",
                    "probability": 0.81201171875
                },
                {
                    "start": 961.42,
                    "end": 961.6,
                    "word": ",",
                    "probability": 0.237060546875
                },
                {
                    "start": 961.6,
                    "end": 961.64,
                    "word": " in",
                    "probability": 0.74365234375
                },
                {
                    "start": 961.64,
                    "end": 961.76,
                    "word": " like",
                    "probability": 0.58154296875
                },
                {
                    "start": 961.84,
                    "end": 962.06,
                    "word": " a machine",
                    "probability": 0.63818359375
                },
                {
                    "start": 962.06,
                    "end": 962.34,
                    "word": " learning",
                    "probability": 0.79541015625
                },
                {
                    "start": 962.34,
                    "end": 962.72,
                    "word": " 101",
                    "probability": 0.6875
                }
            ]
        },
        {
            "id": 157,
            "seek": 96260,
            "start": 963.38,
            "end": 970.2,
            "text": " course, you learn about the bag of words model and about discrete states, Markov models of language.",
            "tokens": [
                50832,
                1164,
                11,
                291,
                1466,
                466,
                264,
                3411,
                295,
                2283,
                2316,
                293,
                466,
                27706,
                4368,
                11,
                3934,
                5179,
                5245,
                295,
                2856,
                13,
                51148
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21893601651702607,
            "compression_ratio": 1.6936936936936937,
            "no_speech_prob": 0.0007786750793457031,
            "words": [
                {
                    "start": 963.38,
                    "end": 963.82,
                    "word": " course",
                    "probability": 0.82080078125
                },
                {
                    "start": 963.82,
                    "end": 964.08,
                    "word": ",",
                    "probability": 0.90478515625
                },
                {
                    "start": 964.08,
                    "end": 964.56,
                    "word": " you",
                    "probability": 0.88671875
                },
                {
                    "start": 964.56,
                    "end": 964.74,
                    "word": " learn",
                    "probability": 0.84521484375
                },
                {
                    "start": 964.74,
                    "end": 964.9,
                    "word": " about",
                    "probability": 0.8203125
                },
                {
                    "start": 964.9,
                    "end": 965.06,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 965.06,
                    "end": 965.26,
                    "word": " bag",
                    "probability": 0.7958984375
                },
                {
                    "start": 965.26,
                    "end": 965.38,
                    "word": " of",
                    "probability": 0.7470703125
                },
                {
                    "start": 965.38,
                    "end": 965.6,
                    "word": " words",
                    "probability": 0.6982421875
                },
                {
                    "start": 965.6,
                    "end": 966.02,
                    "word": " model",
                    "probability": 0.8076171875
                },
                {
                    "start": 966.02,
                    "end": 966.26,
                    "word": " and",
                    "probability": 0.7177734375
                },
                {
                    "start": 966.26,
                    "end": 966.48,
                    "word": " about",
                    "probability": 0.7841796875
                },
                {
                    "start": 966.48,
                    "end": 966.94,
                    "word": " discrete",
                    "probability": 0.6884765625
                },
                {
                    "start": 966.94,
                    "end": 968.38,
                    "word": " states",
                    "probability": 0.74365234375
                },
                {
                    "start": 968.38,
                    "end": 968.62,
                    "word": ",",
                    "probability": 0.81640625
                },
                {
                    "start": 968.62,
                    "end": 968.86,
                    "word": " Markov",
                    "probability": 0.92626953125
                },
                {
                    "start": 968.86,
                    "end": 969.18,
                    "word": " models",
                    "probability": 0.7646484375
                },
                {
                    "start": 969.18,
                    "end": 969.36,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 969.36,
                    "end": 969.76,
                    "word": " language",
                    "probability": 0.666015625
                },
                {
                    "start": 969.76,
                    "end": 970.2,
                    "word": ".",
                    "probability": 0.92578125
                }
            ]
        },
        {
            "id": 158,
            "seek": 96260,
            "start": 970.86,
            "end": 975.6,
            "text": " And I want to talk about these representations as a way of talking about how limited those",
            "tokens": [
                51212,
                400,
                286,
                528,
                281,
                751,
                466,
                613,
                33358,
                382,
                257,
                636,
                295,
                1417,
                466,
                577,
                5567,
                729,
                51436
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21893601651702607,
            "compression_ratio": 1.6936936936936937,
            "no_speech_prob": 0.0007786750793457031,
            "words": [
                {
                    "start": 970.86,
                    "end": 971.3,
                    "word": " And",
                    "probability": 0.8798828125
                },
                {
                    "start": 971.3,
                    "end": 971.38,
                    "word": " I",
                    "probability": 0.99658203125
                },
                {
                    "start": 971.38,
                    "end": 971.46,
                    "word": " want",
                    "probability": 0.3525390625
                },
                {
                    "start": 971.46,
                    "end": 971.54,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 971.54,
                    "end": 971.7,
                    "word": " talk",
                    "probability": 0.81005859375
                },
                {
                    "start": 971.7,
                    "end": 971.9,
                    "word": " about",
                    "probability": 0.82177734375
                },
                {
                    "start": 971.9,
                    "end": 972.3,
                    "word": " these",
                    "probability": 0.708984375
                },
                {
                    "start": 972.3,
                    "end": 973.0,
                    "word": " representations",
                    "probability": 0.7890625
                },
                {
                    "start": 973.0,
                    "end": 973.88,
                    "word": " as",
                    "probability": 0.85009765625
                },
                {
                    "start": 973.98,
                    "end": 974.08,
                    "word": " a way",
                    "probability": 0.9013671875
                },
                {
                    "start": 974.08,
                    "end": 974.2,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 974.2,
                    "end": 974.42,
                    "word": " talking",
                    "probability": 0.76611328125
                },
                {
                    "start": 974.42,
                    "end": 974.64,
                    "word": " about",
                    "probability": 0.8193359375
                },
                {
                    "start": 974.64,
                    "end": 974.82,
                    "word": " how",
                    "probability": 0.88525390625
                },
                {
                    "start": 974.82,
                    "end": 975.2,
                    "word": " limited",
                    "probability": 0.8095703125
                },
                {
                    "start": 975.2,
                    "end": 975.6,
                    "word": " those",
                    "probability": 0.775390625
                }
            ]
        },
        {
            "id": 159,
            "seek": 96260,
            "start": 975.6,
            "end": 979.58,
            "text": " representations were, and then I'll talk about how transformers break us free of some of those",
            "tokens": [
                51436,
                33358,
                645,
                11,
                293,
                550,
                286,
                603,
                751,
                466,
                577,
                4088,
                433,
                1821,
                505,
                1737,
                295,
                512,
                295,
                729,
                51636
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21893601651702607,
            "compression_ratio": 1.6936936936936937,
            "no_speech_prob": 0.0007786750793457031,
            "words": [
                {
                    "start": 975.6,
                    "end": 976.0,
                    "word": " representations",
                    "probability": 0.77294921875
                },
                {
                    "start": 976.0,
                    "end": 976.42,
                    "word": " were",
                    "probability": 0.88134765625
                },
                {
                    "start": 976.42,
                    "end": 976.52,
                    "word": ",",
                    "probability": 0.382568359375
                },
                {
                    "start": 976.52,
                    "end": 976.58,
                    "word": " and",
                    "probability": 0.892578125
                },
                {
                    "start": 976.58,
                    "end": 976.68,
                    "word": " then",
                    "probability": 0.72119140625
                },
                {
                    "start": 976.68,
                    "end": 976.76,
                    "word": " I'll",
                    "probability": 0.962158203125
                },
                {
                    "start": 976.76,
                    "end": 976.96,
                    "word": " talk",
                    "probability": 0.82177734375
                },
                {
                    "start": 976.96,
                    "end": 977.12,
                    "word": " about",
                    "probability": 0.8154296875
                },
                {
                    "start": 977.12,
                    "end": 977.28,
                    "word": " how",
                    "probability": 0.88818359375
                },
                {
                    "start": 977.28,
                    "end": 977.94,
                    "word": " transformers",
                    "probability": 0.77978515625
                },
                {
                    "start": 977.94,
                    "end": 978.8,
                    "word": " break",
                    "probability": 0.94287109375
                },
                {
                    "start": 978.8,
                    "end": 978.94,
                    "word": " us",
                    "probability": 0.8798828125
                },
                {
                    "start": 978.94,
                    "end": 979.12,
                    "word": " free",
                    "probability": 0.86767578125
                },
                {
                    "start": 979.12,
                    "end": 979.24,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 979.24,
                    "end": 979.36,
                    "word": " some",
                    "probability": 0.81689453125
                },
                {
                    "start": 979.36,
                    "end": 979.46,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 979.46,
                    "end": 979.58,
                    "word": " those",
                    "probability": 0.83984375
                }
            ]
        },
        {
            "id": 160,
            "seek": 98602,
            "start": 979.58,
            "end": 985.96,
            "text": " limitations. So, a bag of words model of a document is a way of representing a document",
            "tokens": [
                50364,
                15705,
                13,
                407,
                11,
                257,
                3411,
                295,
                2283,
                2316,
                295,
                257,
                4166,
                307,
                257,
                636,
                295,
                13460,
                257,
                4166,
                50684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23330965350297364,
            "compression_ratio": 1.7511737089201878,
            "no_speech_prob": 0.003322601318359375,
            "words": [
                {
                    "start": 979.58,
                    "end": 979.98,
                    "word": " limitations",
                    "probability": 0.350341796875
                },
                {
                    "start": 979.98,
                    "end": 980.5,
                    "word": ".",
                    "probability": 0.9326171875
                },
                {
                    "start": 980.78,
                    "end": 981.0,
                    "word": " So",
                    "probability": 0.806640625
                },
                {
                    "start": 981.0,
                    "end": 981.5,
                    "word": ",",
                    "probability": 0.2012939453125
                },
                {
                    "start": 981.9,
                    "end": 982.12,
                    "word": " a bag",
                    "probability": 0.76171875
                },
                {
                    "start": 982.12,
                    "end": 982.24,
                    "word": " of",
                    "probability": 0.61572265625
                },
                {
                    "start": 982.24,
                    "end": 982.46,
                    "word": " words",
                    "probability": 0.70751953125
                },
                {
                    "start": 982.46,
                    "end": 982.78,
                    "word": " model",
                    "probability": 0.80322265625
                },
                {
                    "start": 982.78,
                    "end": 983.04,
                    "word": " of",
                    "probability": 0.8984375
                },
                {
                    "start": 983.22,
                    "end": 983.82,
                    "word": " a document",
                    "probability": 0.74169921875
                },
                {
                    "start": 983.82,
                    "end": 984.4,
                    "word": " is",
                    "probability": 0.8857421875
                },
                {
                    "start": 984.54,
                    "end": 984.8,
                    "word": " a way",
                    "probability": 0.8955078125
                },
                {
                    "start": 984.8,
                    "end": 985.06,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 985.06,
                    "end": 985.48,
                    "word": " representing",
                    "probability": 0.69873046875
                },
                {
                    "start": 985.68,
                    "end": 985.96,
                    "word": " a document",
                    "probability": 0.84130859375
                }
            ]
        },
        {
            "id": 161,
            "seek": 98602,
            "start": 985.96,
            "end": 990.28,
            "text": " numerically for the purposes of machine learning, in which we just count up how many times each",
            "tokens": [
                50684,
                7866,
                984,
                337,
                264,
                9932,
                295,
                3479,
                2539,
                11,
                294,
                597,
                321,
                445,
                1207,
                493,
                577,
                867,
                1413,
                1184,
                50904
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23330965350297364,
            "compression_ratio": 1.7511737089201878,
            "no_speech_prob": 0.003322601318359375,
            "words": [
                {
                    "start": 985.96,
                    "end": 986.48,
                    "word": " numerically",
                    "probability": 0.895263671875
                },
                {
                    "start": 986.48,
                    "end": 986.68,
                    "word": " for",
                    "probability": 0.86669921875
                },
                {
                    "start": 986.68,
                    "end": 986.8,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 986.8,
                    "end": 987.06,
                    "word": " purposes",
                    "probability": 0.81103515625
                },
                {
                    "start": 987.06,
                    "end": 987.26,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 987.26,
                    "end": 987.52,
                    "word": " machine",
                    "probability": 0.68896484375
                },
                {
                    "start": 987.52,
                    "end": 987.86,
                    "word": " learning",
                    "probability": 0.84619140625
                },
                {
                    "start": 987.86,
                    "end": 988.06,
                    "word": ",",
                    "probability": 0.466796875
                },
                {
                    "start": 988.06,
                    "end": 988.46,
                    "word": " in",
                    "probability": 0.8896484375
                },
                {
                    "start": 988.46,
                    "end": 988.6,
                    "word": " which",
                    "probability": 0.9150390625
                },
                {
                    "start": 988.6,
                    "end": 988.72,
                    "word": " we",
                    "probability": 0.8974609375
                },
                {
                    "start": 988.72,
                    "end": 988.9,
                    "word": " just",
                    "probability": 0.876953125
                },
                {
                    "start": 988.9,
                    "end": 989.22,
                    "word": " count",
                    "probability": 0.8173828125
                },
                {
                    "start": 989.22,
                    "end": 989.42,
                    "word": " up",
                    "probability": 0.896484375
                },
                {
                    "start": 989.42,
                    "end": 989.54,
                    "word": " how",
                    "probability": 0.8994140625
                },
                {
                    "start": 989.54,
                    "end": 989.68,
                    "word": " many",
                    "probability": 0.80615234375
                },
                {
                    "start": 989.68,
                    "end": 989.96,
                    "word": " times",
                    "probability": 0.81640625
                },
                {
                    "start": 989.96,
                    "end": 990.28,
                    "word": " each",
                    "probability": 0.875
                }
            ]
        },
        {
            "id": 162,
            "seek": 98602,
            "start": 990.28,
            "end": 997.6,
            "text": " word in that document appeared. And then we create a matrix out of all the documents we have,",
            "tokens": [
                50904,
                1349,
                294,
                300,
                4166,
                8516,
                13,
                400,
                550,
                321,
                1884,
                257,
                8141,
                484,
                295,
                439,
                264,
                8512,
                321,
                362,
                11,
                51244
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23330965350297364,
            "compression_ratio": 1.7511737089201878,
            "no_speech_prob": 0.003322601318359375,
            "words": [
                {
                    "start": 990.28,
                    "end": 990.74,
                    "word": " word",
                    "probability": 0.55859375
                },
                {
                    "start": 990.74,
                    "end": 991.1,
                    "word": " in",
                    "probability": 0.892578125
                },
                {
                    "start": 991.1,
                    "end": 991.28,
                    "word": " that",
                    "probability": 0.876953125
                },
                {
                    "start": 991.28,
                    "end": 991.68,
                    "word": " document",
                    "probability": 0.83837890625
                },
                {
                    "start": 991.68,
                    "end": 992.14,
                    "word": " appeared",
                    "probability": 0.72265625
                },
                {
                    "start": 992.14,
                    "end": 992.66,
                    "word": ".",
                    "probability": 0.490234375
                },
                {
                    "start": 993.04,
                    "end": 993.56,
                    "word": " And",
                    "probability": 0.82373046875
                },
                {
                    "start": 993.56,
                    "end": 993.74,
                    "word": " then",
                    "probability": 0.72216796875
                },
                {
                    "start": 993.74,
                    "end": 993.94,
                    "word": " we",
                    "probability": 0.8466796875
                },
                {
                    "start": 993.94,
                    "end": 994.58,
                    "word": " create",
                    "probability": 0.83349609375
                },
                {
                    "start": 995.24,
                    "end": 995.92,
                    "word": " a matrix",
                    "probability": 0.8173828125
                },
                {
                    "start": 995.92,
                    "end": 996.22,
                    "word": " out",
                    "probability": 0.76123046875
                },
                {
                    "start": 996.22,
                    "end": 996.3,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 996.3,
                    "end": 996.4,
                    "word": " all",
                    "probability": 0.90087890625
                },
                {
                    "start": 996.4,
                    "end": 996.48,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 996.48,
                    "end": 996.8,
                    "word": " documents",
                    "probability": 0.79052734375
                },
                {
                    "start": 996.8,
                    "end": 996.98,
                    "word": " we",
                    "probability": 0.875
                },
                {
                    "start": 996.98,
                    "end": 997.28,
                    "word": " have",
                    "probability": 0.88037109375
                },
                {
                    "start": 997.28,
                    "end": 997.6,
                    "word": ",",
                    "probability": 0.45703125
                }
            ]
        },
        {
            "id": 163,
            "seek": 98602,
            "start": 997.6,
            "end": 1003.0,
            "text": " and all the words in all those documents. And the entries in those matrix are just word counts.",
            "tokens": [
                51276,
                293,
                439,
                264,
                2283,
                294,
                439,
                729,
                8512,
                13,
                400,
                264,
                23041,
                294,
                729,
                8141,
                366,
                445,
                1349,
                14893,
                13,
                51516
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23330965350297364,
            "compression_ratio": 1.7511737089201878,
            "no_speech_prob": 0.003322601318359375,
            "words": [
                {
                    "start": 997.6,
                    "end": 998.02,
                    "word": " and",
                    "probability": 0.8505859375
                },
                {
                    "start": 998.02,
                    "end": 998.12,
                    "word": " all",
                    "probability": 0.8974609375
                },
                {
                    "start": 998.12,
                    "end": 998.22,
                    "word": " the",
                    "probability": 0.81884765625
                },
                {
                    "start": 998.22,
                    "end": 998.4,
                    "word": " words",
                    "probability": 0.72216796875
                },
                {
                    "start": 998.4,
                    "end": 998.5,
                    "word": " in",
                    "probability": 0.53271484375
                },
                {
                    "start": 998.5,
                    "end": 998.62,
                    "word": " all",
                    "probability": 0.8798828125
                },
                {
                    "start": 998.62,
                    "end": 998.76,
                    "word": " those",
                    "probability": 0.8349609375
                },
                {
                    "start": 998.76,
                    "end": 999.3,
                    "word": " documents",
                    "probability": 0.796875
                },
                {
                    "start": 999.3,
                    "end": 999.82,
                    "word": ".",
                    "probability": 0.556640625
                },
                {
                    "start": 999.92,
                    "end": 1000.38,
                    "word": " And",
                    "probability": 0.85498046875
                },
                {
                    "start": 1000.38,
                    "end": 1000.46,
                    "word": " the",
                    "probability": 0.55224609375
                },
                {
                    "start": 1000.46,
                    "end": 1000.66,
                    "word": " entries",
                    "probability": 0.78857421875
                },
                {
                    "start": 1000.66,
                    "end": 1000.82,
                    "word": " in",
                    "probability": 0.88427734375
                },
                {
                    "start": 1000.82,
                    "end": 1000.98,
                    "word": " those",
                    "probability": 0.802734375
                },
                {
                    "start": 1000.98,
                    "end": 1001.22,
                    "word": " matrix",
                    "probability": 0.52001953125
                },
                {
                    "start": 1001.22,
                    "end": 1001.42,
                    "word": " are",
                    "probability": 0.84619140625
                },
                {
                    "start": 1001.42,
                    "end": 1001.68,
                    "word": " just",
                    "probability": 0.884765625
                },
                {
                    "start": 1001.68,
                    "end": 1002.06,
                    "word": " word",
                    "probability": 0.79638671875
                },
                {
                    "start": 1002.06,
                    "end": 1002.48,
                    "word": " counts",
                    "probability": 0.744140625
                },
                {
                    "start": 1002.48,
                    "end": 1003.0,
                    "word": ".",
                    "probability": 0.81298828125
                }
            ]
        },
        {
            "id": 164,
            "seek": 101414,
            "start": 1003.0,
            "end": 1010.32,
            "text": " So, in the case on the left over here, the column vectors in our matrix are documents,",
            "tokens": [
                50376,
                407,
                11,
                294,
                264,
                1389,
                322,
                264,
                1411,
                670,
                510,
                11,
                264,
                7738,
                18875,
                294,
                527,
                8141,
                366,
                8512,
                11,
                50724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2156894379055377,
            "compression_ratio": 1.7851239669421488,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 1003.0,
                    "end": 1003.54,
                    "word": " So",
                    "probability": 0.34765625
                },
                {
                    "start": 1003.54,
                    "end": 1004.0,
                    "word": ",",
                    "probability": 0.2587890625
                },
                {
                    "start": 1004.0,
                    "end": 1004.3,
                    "word": " in",
                    "probability": 0.8173828125
                },
                {
                    "start": 1004.3,
                    "end": 1004.82,
                    "word": " the",
                    "probability": 0.73779296875
                },
                {
                    "start": 1004.82,
                    "end": 1005.16,
                    "word": " case",
                    "probability": 0.826171875
                },
                {
                    "start": 1005.16,
                    "end": 1005.9,
                    "word": " on",
                    "probability": 0.82373046875
                },
                {
                    "start": 1005.9,
                    "end": 1006.04,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 1006.04,
                    "end": 1006.2,
                    "word": " left",
                    "probability": 0.8603515625
                },
                {
                    "start": 1006.2,
                    "end": 1006.38,
                    "word": " over",
                    "probability": 0.85888671875
                },
                {
                    "start": 1006.38,
                    "end": 1006.72,
                    "word": " here",
                    "probability": 0.8134765625
                },
                {
                    "start": 1006.72,
                    "end": 1007.3,
                    "word": ",",
                    "probability": 0.82763671875
                },
                {
                    "start": 1007.3,
                    "end": 1007.42,
                    "word": " the",
                    "probability": 0.7890625
                },
                {
                    "start": 1007.42,
                    "end": 1007.86,
                    "word": " column",
                    "probability": 0.81005859375
                },
                {
                    "start": 1007.86,
                    "end": 1008.28,
                    "word": " vectors",
                    "probability": 0.90234375
                },
                {
                    "start": 1008.28,
                    "end": 1008.44,
                    "word": " in",
                    "probability": 0.87646484375
                },
                {
                    "start": 1008.44,
                    "end": 1008.56,
                    "word": " our",
                    "probability": 0.8095703125
                },
                {
                    "start": 1008.56,
                    "end": 1008.86,
                    "word": " matrix",
                    "probability": 0.83349609375
                },
                {
                    "start": 1008.86,
                    "end": 1009.18,
                    "word": " are",
                    "probability": 0.86669921875
                },
                {
                    "start": 1009.18,
                    "end": 1009.86,
                    "word": " documents",
                    "probability": 0.7490234375
                },
                {
                    "start": 1009.86,
                    "end": 1010.32,
                    "word": ",",
                    "probability": 0.314453125
                }
            ]
        },
        {
            "id": 165,
            "seek": 101414,
            "start": 1010.32,
            "end": 1016.16,
            "text": " and they get one dimension per word in our vocabulary, and they get counts of how many",
            "tokens": [
                50724,
                293,
                436,
                483,
                472,
                10139,
                680,
                1349,
                294,
                527,
                19864,
                11,
                293,
                436,
                483,
                14893,
                295,
                577,
                867,
                51016
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2156894379055377,
            "compression_ratio": 1.7851239669421488,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 1010.32,
                    "end": 1010.66,
                    "word": " and",
                    "probability": 0.88623046875
                },
                {
                    "start": 1010.66,
                    "end": 1011.88,
                    "word": " they",
                    "probability": 0.6943359375
                },
                {
                    "start": 1011.88,
                    "end": 1012.08,
                    "word": " get",
                    "probability": 0.90087890625
                },
                {
                    "start": 1012.08,
                    "end": 1012.34,
                    "word": " one",
                    "probability": 0.78271484375
                },
                {
                    "start": 1012.34,
                    "end": 1012.76,
                    "word": " dimension",
                    "probability": 0.9208984375
                },
                {
                    "start": 1012.76,
                    "end": 1013.36,
                    "word": " per",
                    "probability": 0.90771484375
                },
                {
                    "start": 1013.36,
                    "end": 1013.92,
                    "word": " word",
                    "probability": 0.7578125
                },
                {
                    "start": 1013.92,
                    "end": 1014.14,
                    "word": " in",
                    "probability": 0.85595703125
                },
                {
                    "start": 1014.14,
                    "end": 1014.26,
                    "word": " our",
                    "probability": 0.78173828125
                },
                {
                    "start": 1014.26,
                    "end": 1014.64,
                    "word": " vocabulary",
                    "probability": 0.681640625
                },
                {
                    "start": 1014.64,
                    "end": 1015.06,
                    "word": ",",
                    "probability": 0.68359375
                },
                {
                    "start": 1015.06,
                    "end": 1015.16,
                    "word": " and",
                    "probability": 0.8876953125
                },
                {
                    "start": 1015.16,
                    "end": 1015.28,
                    "word": " they",
                    "probability": 0.73046875
                },
                {
                    "start": 1015.28,
                    "end": 1015.44,
                    "word": " get",
                    "probability": 0.89697265625
                },
                {
                    "start": 1015.44,
                    "end": 1015.76,
                    "word": " counts",
                    "probability": 0.6875
                },
                {
                    "start": 1015.76,
                    "end": 1015.9,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 1015.9,
                    "end": 1016.02,
                    "word": " how",
                    "probability": 0.89404296875
                },
                {
                    "start": 1016.02,
                    "end": 1016.16,
                    "word": " many",
                    "probability": 0.80908203125
                }
            ]
        },
        {
            "id": 166,
            "seek": 101414,
            "start": 1016.16,
            "end": 1021.56,
            "text": " times words appeared in that vocabulary in the entries of that particular vector.",
            "tokens": [
                51016,
                1413,
                2283,
                8516,
                294,
                300,
                19864,
                294,
                264,
                23041,
                295,
                300,
                1729,
                8062,
                13,
                51280
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2156894379055377,
            "compression_ratio": 1.7851239669421488,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 1016.16,
                    "end": 1016.4,
                    "word": " times",
                    "probability": 0.8046875
                },
                {
                    "start": 1016.4,
                    "end": 1016.62,
                    "word": " words",
                    "probability": 0.69921875
                },
                {
                    "start": 1016.62,
                    "end": 1017.06,
                    "word": " appeared",
                    "probability": 0.68701171875
                },
                {
                    "start": 1017.06,
                    "end": 1018.58,
                    "word": " in",
                    "probability": 0.83447265625
                },
                {
                    "start": 1018.58,
                    "end": 1018.74,
                    "word": " that",
                    "probability": 0.86572265625
                },
                {
                    "start": 1018.74,
                    "end": 1019.14,
                    "word": " vocabulary",
                    "probability": 0.72021484375
                },
                {
                    "start": 1019.14,
                    "end": 1019.58,
                    "word": " in",
                    "probability": 0.51220703125
                },
                {
                    "start": 1019.58,
                    "end": 1019.74,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 1019.74,
                    "end": 1020.06,
                    "word": " entries",
                    "probability": 0.7841796875
                },
                {
                    "start": 1020.06,
                    "end": 1020.34,
                    "word": " of",
                    "probability": 0.89990234375
                },
                {
                    "start": 1020.34,
                    "end": 1020.62,
                    "word": " that",
                    "probability": 0.8603515625
                },
                {
                    "start": 1020.62,
                    "end": 1020.94,
                    "word": " particular",
                    "probability": 0.79931640625
                },
                {
                    "start": 1020.94,
                    "end": 1021.26,
                    "word": " vector",
                    "probability": 0.90234375
                },
                {
                    "start": 1021.26,
                    "end": 1021.56,
                    "word": ".",
                    "probability": 0.96875
                }
            ]
        },
        {
            "id": 167,
            "seek": 101414,
            "start": 1022.84,
            "end": 1027.12,
            "text": " And hopefully it's intuitive that, you know, once you've represented your documents in a",
            "tokens": [
                51348,
                400,
                4696,
                309,
                311,
                21769,
                300,
                11,
                291,
                458,
                11,
                1564,
                291,
                600,
                10379,
                428,
                8512,
                294,
                257,
                51544
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2156894379055377,
            "compression_ratio": 1.7851239669421488,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 1022.84,
                    "end": 1022.98,
                    "word": " And",
                    "probability": 0.603515625
                },
                {
                    "start": 1022.98,
                    "end": 1023.42,
                    "word": " hopefully",
                    "probability": 0.54296875
                },
                {
                    "start": 1023.42,
                    "end": 1023.56,
                    "word": " it's",
                    "probability": 0.829833984375
                },
                {
                    "start": 1023.56,
                    "end": 1023.86,
                    "word": " intuitive",
                    "probability": 0.7822265625
                },
                {
                    "start": 1023.86,
                    "end": 1024.22,
                    "word": " that",
                    "probability": 0.84228515625
                },
                {
                    "start": 1024.22,
                    "end": 1024.28,
                    "word": ",",
                    "probability": 0.64111328125
                },
                {
                    "start": 1024.28,
                    "end": 1024.38,
                    "word": " you",
                    "probability": 0.82373046875
                },
                {
                    "start": 1024.38,
                    "end": 1024.5,
                    "word": " know",
                    "probability": 0.81298828125
                },
                {
                    "start": 1024.5,
                    "end": 1024.86,
                    "word": ",",
                    "probability": 0.98681640625
                },
                {
                    "start": 1024.86,
                    "end": 1025.08,
                    "word": " once",
                    "probability": 0.8193359375
                },
                {
                    "start": 1025.08,
                    "end": 1025.26,
                    "word": " you've",
                    "probability": 0.9091796875
                },
                {
                    "start": 1025.26,
                    "end": 1025.54,
                    "word": " represented",
                    "probability": 0.720703125
                },
                {
                    "start": 1025.54,
                    "end": 1025.74,
                    "word": " your",
                    "probability": 0.80712890625
                },
                {
                    "start": 1025.74,
                    "end": 1026.1,
                    "word": " documents",
                    "probability": 0.79833984375
                },
                {
                    "start": 1026.1,
                    "end": 1026.48,
                    "word": " in",
                    "probability": 0.8134765625
                },
                {
                    "start": 1026.76,
                    "end": 1027.12,
                    "word": " a vocabulary",
                    "probability": 0.71435546875
                }
            ]
        },
        {
            "id": 168,
            "seek": 101414,
            "start": 1027.12,
            "end": 1031.12,
            "text": " vocabulary space in this way, you can compare documents by taking some distance measure",
            "tokens": [
                51544,
                19864,
                1901,
                294,
                341,
                636,
                11,
                291,
                393,
                6794,
                8512,
                538,
                1940,
                512,
                4560,
                3481,
                51792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2156894379055377,
            "compression_ratio": 1.7851239669421488,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 1027.12,
                    "end": 1027.58,
                    "word": " space",
                    "probability": 0.82373046875
                },
                {
                    "start": 1027.58,
                    "end": 1027.72,
                    "word": " in",
                    "probability": 0.77587890625
                },
                {
                    "start": 1027.72,
                    "end": 1027.86,
                    "word": " this",
                    "probability": 0.8798828125
                },
                {
                    "start": 1027.86,
                    "end": 1028.04,
                    "word": " way",
                    "probability": 0.89990234375
                },
                {
                    "start": 1028.04,
                    "end": 1028.08,
                    "word": ",",
                    "probability": 0.7841796875
                },
                {
                    "start": 1028.08,
                    "end": 1028.2,
                    "word": " you",
                    "probability": 0.89111328125
                },
                {
                    "start": 1028.2,
                    "end": 1028.34,
                    "word": " can",
                    "probability": 0.89990234375
                },
                {
                    "start": 1028.34,
                    "end": 1028.62,
                    "word": " compare",
                    "probability": 0.86474609375
                },
                {
                    "start": 1028.62,
                    "end": 1029.18,
                    "word": " documents",
                    "probability": 0.794921875
                },
                {
                    "start": 1029.18,
                    "end": 1029.58,
                    "word": " by",
                    "probability": 0.8984375
                },
                {
                    "start": 1029.58,
                    "end": 1030.16,
                    "word": " taking",
                    "probability": 0.81396484375
                },
                {
                    "start": 1030.16,
                    "end": 1030.4,
                    "word": " some",
                    "probability": 0.83056640625
                },
                {
                    "start": 1030.4,
                    "end": 1030.7,
                    "word": " distance",
                    "probability": 0.79736328125
                },
                {
                    "start": 1030.7,
                    "end": 1031.12,
                    "word": " measure",
                    "probability": 0.7802734375
                }
            ]
        },
        {
            "id": 169,
            "seek": 104178,
            "start": 1031.12,
            "end": 1036.4,
            "text": " between pairs of documents. You can also train machine learning models on your document corpus",
            "tokens": [
                50384,
                1296,
                15494,
                295,
                8512,
                13,
                509,
                393,
                611,
                3847,
                3479,
                2539,
                5245,
                322,
                428,
                4166,
                1181,
                31624,
                50628
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23723322462017943,
            "compression_ratio": 1.65625,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1031.12,
                    "end": 1031.84,
                    "word": " between",
                    "probability": 0.329833984375
                },
                {
                    "start": 1031.84,
                    "end": 1032.16,
                    "word": " pairs",
                    "probability": 0.6455078125
                },
                {
                    "start": 1032.16,
                    "end": 1032.3,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1032.3,
                    "end": 1032.72,
                    "word": " documents",
                    "probability": 0.7939453125
                },
                {
                    "start": 1032.72,
                    "end": 1033.2,
                    "word": ".",
                    "probability": 0.72119140625
                },
                {
                    "start": 1033.2,
                    "end": 1033.24,
                    "word": " You",
                    "probability": 0.87255859375
                },
                {
                    "start": 1033.24,
                    "end": 1033.38,
                    "word": " can",
                    "probability": 0.89404296875
                },
                {
                    "start": 1033.38,
                    "end": 1033.6,
                    "word": " also",
                    "probability": 0.8251953125
                },
                {
                    "start": 1033.6,
                    "end": 1034.0,
                    "word": " train",
                    "probability": 0.8369140625
                },
                {
                    "start": 1034.0,
                    "end": 1034.26,
                    "word": " machine",
                    "probability": 0.68212890625
                },
                {
                    "start": 1034.26,
                    "end": 1034.5,
                    "word": " learning",
                    "probability": 0.80908203125
                },
                {
                    "start": 1034.5,
                    "end": 1034.9,
                    "word": " models",
                    "probability": 0.7861328125
                },
                {
                    "start": 1034.9,
                    "end": 1035.62,
                    "word": " on",
                    "probability": 0.8720703125
                },
                {
                    "start": 1035.62,
                    "end": 1035.74,
                    "word": " your",
                    "probability": 0.80517578125
                },
                {
                    "start": 1035.74,
                    "end": 1036.04,
                    "word": " document",
                    "probability": 0.7978515625
                },
                {
                    "start": 1036.04,
                    "end": 1036.4,
                    "word": " corpus",
                    "probability": 0.90673828125
                }
            ]
        },
        {
            "id": 170,
            "seek": 104178,
            "start": 1036.4,
            "end": 1044.52,
            "text": " to say classified news articles is about sports or politics. But what you've done in the first",
            "tokens": [
                50628,
                281,
                584,
                20627,
                2583,
                11290,
                307,
                466,
                6573,
                420,
                7341,
                13,
                583,
                437,
                291,
                600,
                1096,
                294,
                264,
                700,
                51032
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23723322462017943,
            "compression_ratio": 1.65625,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1036.4,
                    "end": 1036.58,
                    "word": " to",
                    "probability": 0.85107421875
                },
                {
                    "start": 1036.58,
                    "end": 1036.76,
                    "word": " say",
                    "probability": 0.767578125
                },
                {
                    "start": 1036.76,
                    "end": 1037.24,
                    "word": " classified",
                    "probability": 0.25146484375
                },
                {
                    "start": 1037.24,
                    "end": 1037.54,
                    "word": " news",
                    "probability": 0.74853515625
                },
                {
                    "start": 1037.54,
                    "end": 1037.86,
                    "word": " articles",
                    "probability": 0.82666015625
                },
                {
                    "start": 1037.86,
                    "end": 1038.16,
                    "word": " is",
                    "probability": 0.64599609375
                },
                {
                    "start": 1038.16,
                    "end": 1038.52,
                    "word": " about",
                    "probability": 0.79296875
                },
                {
                    "start": 1038.52,
                    "end": 1039.16,
                    "word": " sports",
                    "probability": 0.75732421875
                },
                {
                    "start": 1039.16,
                    "end": 1039.5,
                    "word": " or",
                    "probability": 0.88134765625
                },
                {
                    "start": 1039.5,
                    "end": 1040.34,
                    "word": " politics",
                    "probability": 0.7421875
                },
                {
                    "start": 1040.34,
                    "end": 1040.96,
                    "word": ".",
                    "probability": 0.8193359375
                },
                {
                    "start": 1041.8,
                    "end": 1042.22,
                    "word": " But",
                    "probability": 0.78369140625
                },
                {
                    "start": 1042.22,
                    "end": 1043.7,
                    "word": " what",
                    "probability": 0.7666015625
                },
                {
                    "start": 1043.7,
                    "end": 1043.88,
                    "word": " you've",
                    "probability": 0.911865234375
                },
                {
                    "start": 1043.88,
                    "end": 1044.04,
                    "word": " done",
                    "probability": 0.8134765625
                },
                {
                    "start": 1044.04,
                    "end": 1044.14,
                    "word": " in",
                    "probability": 0.7451171875
                },
                {
                    "start": 1044.14,
                    "end": 1044.26,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1044.26,
                    "end": 1044.52,
                    "word": " first",
                    "probability": 0.81103515625
                }
            ]
        },
        {
            "id": 171,
            "seek": 104178,
            "start": 1044.52,
            "end": 1052.14,
            "text": " step of these models is drop out sequence information. So, you've forgotten about which",
            "tokens": [
                51032,
                1823,
                295,
                613,
                5245,
                307,
                3270,
                484,
                8310,
                1589,
                13,
                407,
                11,
                291,
                600,
                11832,
                466,
                597,
                51412
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23723322462017943,
            "compression_ratio": 1.65625,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1044.52,
                    "end": 1044.9,
                    "word": " step",
                    "probability": 0.802734375
                },
                {
                    "start": 1044.9,
                    "end": 1045.32,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 1045.32,
                    "end": 1045.82,
                    "word": " these",
                    "probability": 0.63671875
                },
                {
                    "start": 1045.82,
                    "end": 1046.22,
                    "word": " models",
                    "probability": 0.78955078125
                },
                {
                    "start": 1046.22,
                    "end": 1047.18,
                    "word": " is",
                    "probability": 0.869140625
                },
                {
                    "start": 1047.18,
                    "end": 1048.76,
                    "word": " drop",
                    "probability": 0.72021484375
                },
                {
                    "start": 1048.76,
                    "end": 1049.06,
                    "word": " out",
                    "probability": 0.7763671875
                },
                {
                    "start": 1049.06,
                    "end": 1049.5,
                    "word": " sequence",
                    "probability": 0.8955078125
                },
                {
                    "start": 1049.5,
                    "end": 1050.04,
                    "word": " information",
                    "probability": 0.73193359375
                },
                {
                    "start": 1050.04,
                    "end": 1050.44,
                    "word": ".",
                    "probability": 0.6201171875
                },
                {
                    "start": 1050.44,
                    "end": 1050.72,
                    "word": " So",
                    "probability": 0.7509765625
                },
                {
                    "start": 1050.72,
                    "end": 1050.88,
                    "word": ",",
                    "probability": 0.0826416015625
                },
                {
                    "start": 1050.88,
                    "end": 1051.16,
                    "word": " you've",
                    "probability": 0.790771484375
                },
                {
                    "start": 1051.16,
                    "end": 1051.46,
                    "word": " forgotten",
                    "probability": 0.7421875
                },
                {
                    "start": 1051.46,
                    "end": 1051.84,
                    "word": " about",
                    "probability": 0.794921875
                },
                {
                    "start": 1051.84,
                    "end": 1052.14,
                    "word": " which",
                    "probability": 0.91845703125
                }
            ]
        },
        {
            "id": 172,
            "seek": 104178,
            "start": 1052.14,
            "end": 1058.76,
            "text": " words come in which order in the document, and you've just represented your document as a bag",
            "tokens": [
                51412,
                2283,
                808,
                294,
                597,
                1668,
                294,
                264,
                4166,
                11,
                293,
                291,
                600,
                445,
                10379,
                428,
                4166,
                382,
                257,
                3411,
                51744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23723322462017943,
            "compression_ratio": 1.65625,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1052.14,
                    "end": 1052.62,
                    "word": " words",
                    "probability": 0.7119140625
                },
                {
                    "start": 1052.62,
                    "end": 1053.64,
                    "word": " come",
                    "probability": 0.7822265625
                },
                {
                    "start": 1053.64,
                    "end": 1053.76,
                    "word": " in",
                    "probability": 0.90234375
                },
                {
                    "start": 1053.76,
                    "end": 1053.9,
                    "word": " which",
                    "probability": 0.912109375
                },
                {
                    "start": 1053.9,
                    "end": 1054.14,
                    "word": " order",
                    "probability": 0.82080078125
                },
                {
                    "start": 1054.14,
                    "end": 1054.28,
                    "word": " in",
                    "probability": 0.87744140625
                },
                {
                    "start": 1054.28,
                    "end": 1054.36,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 1054.36,
                    "end": 1054.72,
                    "word": " document",
                    "probability": 0.837890625
                },
                {
                    "start": 1054.72,
                    "end": 1055.52,
                    "word": ",",
                    "probability": 0.266357421875
                },
                {
                    "start": 1055.52,
                    "end": 1056.82,
                    "word": " and",
                    "probability": 0.890625
                },
                {
                    "start": 1056.82,
                    "end": 1057.16,
                    "word": " you've",
                    "probability": 0.887451171875
                },
                {
                    "start": 1057.16,
                    "end": 1057.3,
                    "word": " just",
                    "probability": 0.8671875
                },
                {
                    "start": 1057.3,
                    "end": 1057.72,
                    "word": " represented",
                    "probability": 0.7109375
                },
                {
                    "start": 1057.72,
                    "end": 1057.92,
                    "word": " your",
                    "probability": 0.779296875
                },
                {
                    "start": 1057.92,
                    "end": 1058.22,
                    "word": " document",
                    "probability": 0.82666015625
                },
                {
                    "start": 1058.22,
                    "end": 1058.48,
                    "word": " as",
                    "probability": 0.90478515625
                },
                {
                    "start": 1058.6,
                    "end": 1058.76,
                    "word": " a bag",
                    "probability": 0.90283203125
                }
            ]
        },
        {
            "id": 173,
            "seek": 106818,
            "start": 1058.76,
            "end": 1063.82,
            "text": " of words. Which is a useful simplifying assumption, and it's one that we still use",
            "tokens": [
                50364,
                295,
                2283,
                13,
                3013,
                307,
                257,
                4420,
                6883,
                5489,
                15302,
                11,
                293,
                309,
                311,
                472,
                300,
                321,
                920,
                764,
                50612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21596534712479845,
            "compression_ratio": 1.6625,
            "no_speech_prob": 0.0013036727905273438,
            "words": [
                {
                    "start": 1058.76,
                    "end": 1058.88,
                    "word": " of",
                    "probability": 0.419189453125
                },
                {
                    "start": 1058.88,
                    "end": 1059.24,
                    "word": " words",
                    "probability": 0.6494140625
                },
                {
                    "start": 1059.24,
                    "end": 1059.6,
                    "word": ".",
                    "probability": 0.08160400390625
                },
                {
                    "start": 1059.64,
                    "end": 1060.16,
                    "word": " Which",
                    "probability": 0.69140625
                },
                {
                    "start": 1060.16,
                    "end": 1060.26,
                    "word": " is",
                    "probability": 0.90087890625
                },
                {
                    "start": 1060.32,
                    "end": 1060.52,
                    "word": " a useful",
                    "probability": 0.7294921875
                },
                {
                    "start": 1060.52,
                    "end": 1060.98,
                    "word": " simplifying",
                    "probability": 0.912841796875
                },
                {
                    "start": 1060.98,
                    "end": 1061.36,
                    "word": " assumption",
                    "probability": 0.8876953125
                },
                {
                    "start": 1061.36,
                    "end": 1061.7,
                    "word": ",",
                    "probability": 0.319091796875
                },
                {
                    "start": 1061.7,
                    "end": 1061.86,
                    "word": " and",
                    "probability": 0.86474609375
                },
                {
                    "start": 1061.86,
                    "end": 1062.06,
                    "word": " it's",
                    "probability": 0.8349609375
                },
                {
                    "start": 1062.06,
                    "end": 1062.22,
                    "word": " one",
                    "probability": 0.806640625
                },
                {
                    "start": 1062.22,
                    "end": 1062.52,
                    "word": " that",
                    "probability": 0.841796875
                },
                {
                    "start": 1062.52,
                    "end": 1063.46,
                    "word": " we",
                    "probability": 0.86767578125
                },
                {
                    "start": 1063.46,
                    "end": 1063.64,
                    "word": " still",
                    "probability": 0.88623046875
                },
                {
                    "start": 1063.64,
                    "end": 1063.82,
                    "word": " use",
                    "probability": 0.81103515625
                }
            ]
        },
        {
            "id": 174,
            "seek": 106818,
            "start": 1063.82,
            "end": 1068.72,
            "text": " today in some of the modeling we do in my research group at Sophos. But it throws out",
            "tokens": [
                50612,
                965,
                294,
                512,
                295,
                264,
                15983,
                321,
                360,
                294,
                452,
                2132,
                1594,
                412,
                18921,
                329,
                13,
                583,
                309,
                19251,
                484,
                50860
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21596534712479845,
            "compression_ratio": 1.6625,
            "no_speech_prob": 0.0013036727905273438,
            "words": [
                {
                    "start": 1063.82,
                    "end": 1064.0,
                    "word": " today",
                    "probability": 0.76806640625
                },
                {
                    "start": 1064.0,
                    "end": 1064.12,
                    "word": " in",
                    "probability": 0.85888671875
                },
                {
                    "start": 1064.12,
                    "end": 1064.22,
                    "word": " some",
                    "probability": 0.79296875
                },
                {
                    "start": 1064.22,
                    "end": 1064.28,
                    "word": " of",
                    "probability": 0.89794921875
                },
                {
                    "start": 1064.28,
                    "end": 1064.36,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 1064.36,
                    "end": 1064.54,
                    "word": " modeling",
                    "probability": 0.611328125
                },
                {
                    "start": 1064.54,
                    "end": 1064.68,
                    "word": " we",
                    "probability": 0.80908203125
                },
                {
                    "start": 1064.68,
                    "end": 1064.84,
                    "word": " do",
                    "probability": 0.88525390625
                },
                {
                    "start": 1064.84,
                    "end": 1065.16,
                    "word": " in",
                    "probability": 0.76171875
                },
                {
                    "start": 1065.16,
                    "end": 1065.6,
                    "word": " my",
                    "probability": 0.8525390625
                },
                {
                    "start": 1065.6,
                    "end": 1065.88,
                    "word": " research",
                    "probability": 0.77001953125
                },
                {
                    "start": 1065.88,
                    "end": 1066.1,
                    "word": " group",
                    "probability": 0.849609375
                },
                {
                    "start": 1066.1,
                    "end": 1066.44,
                    "word": " at",
                    "probability": 0.87158203125
                },
                {
                    "start": 1066.44,
                    "end": 1066.9,
                    "word": " Sophos",
                    "probability": 0.764404296875
                },
                {
                    "start": 1066.9,
                    "end": 1067.42,
                    "word": ".",
                    "probability": 0.72509765625
                },
                {
                    "start": 1067.72,
                    "end": 1068.16,
                    "word": " But",
                    "probability": 0.8583984375
                },
                {
                    "start": 1068.16,
                    "end": 1068.34,
                    "word": " it",
                    "probability": 0.875
                },
                {
                    "start": 1068.34,
                    "end": 1068.56,
                    "word": " throws",
                    "probability": 0.7900390625
                },
                {
                    "start": 1068.56,
                    "end": 1068.72,
                    "word": " out",
                    "probability": 0.81884765625
                }
            ]
        },
        {
            "id": 175,
            "seek": 106818,
            "start": 1068.88,
            "end": 1072.82,
            "text": " a ton of information that transformers and more modern models don't throw out.",
            "tokens": [
                50860,
                257,
                2952,
                295,
                1589,
                300,
                4088,
                433,
                293,
                544,
                4363,
                5245,
                500,
                380,
                3507,
                484,
                13,
                51036
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21596534712479845,
            "compression_ratio": 1.6625,
            "no_speech_prob": 0.0013036727905273438,
            "words": [
                {
                    "start": 1068.88,
                    "end": 1069.14,
                    "word": " a ton",
                    "probability": 0.884765625
                },
                {
                    "start": 1069.14,
                    "end": 1069.26,
                    "word": " of",
                    "probability": 0.8935546875
                },
                {
                    "start": 1069.26,
                    "end": 1069.64,
                    "word": " information",
                    "probability": 0.748046875
                },
                {
                    "start": 1069.64,
                    "end": 1069.86,
                    "word": " that",
                    "probability": 0.8603515625
                },
                {
                    "start": 1069.86,
                    "end": 1070.52,
                    "word": " transformers",
                    "probability": 0.846435546875
                },
                {
                    "start": 1070.52,
                    "end": 1070.72,
                    "word": " and",
                    "probability": 0.7568359375
                },
                {
                    "start": 1070.72,
                    "end": 1070.88,
                    "word": " more",
                    "probability": 0.87060546875
                },
                {
                    "start": 1070.88,
                    "end": 1071.2,
                    "word": " modern",
                    "probability": 0.833984375
                },
                {
                    "start": 1071.2,
                    "end": 1071.6,
                    "word": " models",
                    "probability": 0.80224609375
                },
                {
                    "start": 1071.6,
                    "end": 1071.82,
                    "word": " don't",
                    "probability": 0.9404296875
                },
                {
                    "start": 1071.82,
                    "end": 1072.02,
                    "word": " throw",
                    "probability": 0.7890625
                },
                {
                    "start": 1072.02,
                    "end": 1072.3,
                    "word": " out",
                    "probability": 0.82177734375
                },
                {
                    "start": 1072.3,
                    "end": 1072.82,
                    "word": ".",
                    "probability": 0.98876953125
                }
            ]
        },
        {
            "id": 176,
            "seek": 106818,
            "start": 1073.74,
            "end": 1078.46,
            "text": " In the model on the right, we have a discrete state Markov model of language.",
            "tokens": [
                51116,
                682,
                264,
                2316,
                322,
                264,
                558,
                11,
                321,
                362,
                257,
                27706,
                1785,
                3934,
                5179,
                2316,
                295,
                2856,
                13,
                51324
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21596534712479845,
            "compression_ratio": 1.6625,
            "no_speech_prob": 0.0013036727905273438,
            "words": [
                {
                    "start": 1073.74,
                    "end": 1074.08,
                    "word": " In",
                    "probability": 0.74560546875
                },
                {
                    "start": 1074.08,
                    "end": 1074.16,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 1074.16,
                    "end": 1074.3,
                    "word": " model",
                    "probability": 0.82373046875
                },
                {
                    "start": 1074.3,
                    "end": 1074.44,
                    "word": " on",
                    "probability": 0.89990234375
                },
                {
                    "start": 1074.44,
                    "end": 1074.52,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 1074.52,
                    "end": 1074.72,
                    "word": " right",
                    "probability": 0.88134765625
                },
                {
                    "start": 1074.72,
                    "end": 1074.96,
                    "word": ",",
                    "probability": 0.67236328125
                },
                {
                    "start": 1074.96,
                    "end": 1075.44,
                    "word": " we",
                    "probability": 0.83251953125
                },
                {
                    "start": 1075.44,
                    "end": 1075.58,
                    "word": " have",
                    "probability": 0.87939453125
                },
                {
                    "start": 1075.68,
                    "end": 1075.96,
                    "word": " a discrete",
                    "probability": 0.787109375
                },
                {
                    "start": 1075.96,
                    "end": 1076.28,
                    "word": " state",
                    "probability": 0.76123046875
                },
                {
                    "start": 1076.28,
                    "end": 1076.7,
                    "word": " Markov",
                    "probability": 0.826171875
                },
                {
                    "start": 1076.7,
                    "end": 1077.0,
                    "word": " model",
                    "probability": 0.8203125
                },
                {
                    "start": 1077.0,
                    "end": 1077.5,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 1077.5,
                    "end": 1077.94,
                    "word": " language",
                    "probability": 0.65576171875
                },
                {
                    "start": 1077.94,
                    "end": 1078.46,
                    "word": ".",
                    "probability": 0.90087890625
                }
            ]
        },
        {
            "id": 177,
            "seek": 106818,
            "start": 1078.74,
            "end": 1085.16,
            "text": " Here each word is a state, and the concept of language given by the model",
            "tokens": [
                51368,
                1692,
                1184,
                1349,
                307,
                257,
                1785,
                11,
                293,
                264,
                3410,
                295,
                2856,
                2212,
                538,
                264,
                2316,
                51680
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21596534712479845,
            "compression_ratio": 1.6625,
            "no_speech_prob": 0.0013036727905273438,
            "words": [
                {
                    "start": 1078.74,
                    "end": 1079.1,
                    "word": " Here",
                    "probability": 0.8095703125
                },
                {
                    "start": 1079.1,
                    "end": 1079.36,
                    "word": " each",
                    "probability": 0.51171875
                },
                {
                    "start": 1079.36,
                    "end": 1079.74,
                    "word": " word",
                    "probability": 0.8134765625
                },
                {
                    "start": 1079.74,
                    "end": 1080.0,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 1080.16,
                    "end": 1080.5,
                    "word": " a state",
                    "probability": 0.8759765625
                },
                {
                    "start": 1080.5,
                    "end": 1081.0,
                    "word": ",",
                    "probability": 0.458251953125
                },
                {
                    "start": 1081.0,
                    "end": 1081.32,
                    "word": " and",
                    "probability": 0.8935546875
                },
                {
                    "start": 1081.32,
                    "end": 1083.08,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 1083.08,
                    "end": 1083.76,
                    "word": " concept",
                    "probability": 0.79638671875
                },
                {
                    "start": 1083.76,
                    "end": 1083.98,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 1083.98,
                    "end": 1084.3,
                    "word": " language",
                    "probability": 0.66748046875
                },
                {
                    "start": 1084.3,
                    "end": 1084.56,
                    "word": " given",
                    "probability": 0.771484375
                },
                {
                    "start": 1084.56,
                    "end": 1084.72,
                    "word": " by",
                    "probability": 0.90380859375
                },
                {
                    "start": 1084.72,
                    "end": 1084.84,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 1084.84,
                    "end": 1085.16,
                    "word": " model",
                    "probability": 0.796875
                }
            ]
        },
        {
            "id": 178,
            "seek": 109818,
            "start": 1086.44,
            "end": 1092.56,
            "text": " is kind of a cheesier adventure story in which the next words in an utterance depends only on",
            "tokens": [
                50388,
                307,
                733,
                295,
                257,
                947,
                279,
                811,
                9868,
                1657,
                294,
                597,
                264,
                958,
                2283,
                294,
                364,
                17567,
                719,
                5946,
                787,
                322,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2641782343939499,
            "compression_ratio": 1.671280276816609,
            "no_speech_prob": 0.0002779960632324219,
            "words": [
                {
                    "start": 1086.44,
                    "end": 1086.96,
                    "word": " is",
                    "probability": 0.72705078125
                },
                {
                    "start": 1086.96,
                    "end": 1087.18,
                    "word": " kind",
                    "probability": 0.69287109375
                },
                {
                    "start": 1087.18,
                    "end": 1087.24,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 1087.36,
                    "end": 1087.64,
                    "word": " a cheesier",
                    "probability": 0.7351888020833334
                },
                {
                    "start": 1087.64,
                    "end": 1087.98,
                    "word": " adventure",
                    "probability": 0.444580078125
                },
                {
                    "start": 1087.98,
                    "end": 1088.38,
                    "word": " story",
                    "probability": 0.87158203125
                },
                {
                    "start": 1088.38,
                    "end": 1088.52,
                    "word": " in",
                    "probability": 0.7841796875
                },
                {
                    "start": 1088.52,
                    "end": 1088.72,
                    "word": " which",
                    "probability": 0.91259765625
                },
                {
                    "start": 1088.72,
                    "end": 1088.9,
                    "word": " the",
                    "probability": 0.728515625
                },
                {
                    "start": 1088.9,
                    "end": 1090.0,
                    "word": " next",
                    "probability": 0.7470703125
                },
                {
                    "start": 1090.0,
                    "end": 1090.54,
                    "word": " words",
                    "probability": 0.53173828125
                },
                {
                    "start": 1090.54,
                    "end": 1091.0,
                    "word": " in",
                    "probability": 0.73095703125
                },
                {
                    "start": 1091.0,
                    "end": 1091.12,
                    "word": " an",
                    "probability": 0.708984375
                },
                {
                    "start": 1091.12,
                    "end": 1091.58,
                    "word": " utterance",
                    "probability": 0.84716796875
                },
                {
                    "start": 1091.58,
                    "end": 1092.06,
                    "word": " depends",
                    "probability": 0.71240234375
                },
                {
                    "start": 1092.06,
                    "end": 1092.38,
                    "word": " only",
                    "probability": 0.82861328125
                },
                {
                    "start": 1092.38,
                    "end": 1092.56,
                    "word": " on",
                    "probability": 0.9111328125
                }
            ]
        },
        {
            "id": 179,
            "seek": 109818,
            "start": 1092.56,
            "end": 1097.62,
            "text": " the current words, and you're just sort of drawing from a probability distribution and moving through",
            "tokens": [
                50728,
                264,
                2190,
                2283,
                11,
                293,
                291,
                434,
                445,
                1333,
                295,
                6316,
                490,
                257,
                8482,
                7316,
                293,
                2684,
                807,
                50984
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2641782343939499,
            "compression_ratio": 1.671280276816609,
            "no_speech_prob": 0.0002779960632324219,
            "words": [
                {
                    "start": 1092.56,
                    "end": 1092.68,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 1092.68,
                    "end": 1092.92,
                    "word": " current",
                    "probability": 0.77490234375
                },
                {
                    "start": 1092.92,
                    "end": 1093.32,
                    "word": " words",
                    "probability": 0.7275390625
                },
                {
                    "start": 1093.32,
                    "end": 1094.08,
                    "word": ",",
                    "probability": 0.227294921875
                },
                {
                    "start": 1094.08,
                    "end": 1094.4,
                    "word": " and",
                    "probability": 0.89306640625
                },
                {
                    "start": 1094.4,
                    "end": 1094.98,
                    "word": " you're",
                    "probability": 0.8984375
                },
                {
                    "start": 1094.98,
                    "end": 1095.1,
                    "word": " just",
                    "probability": 0.8828125
                },
                {
                    "start": 1095.1,
                    "end": 1095.22,
                    "word": " sort",
                    "probability": 0.6875
                },
                {
                    "start": 1095.22,
                    "end": 1095.3,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1095.3,
                    "end": 1095.62,
                    "word": " drawing",
                    "probability": 0.7138671875
                },
                {
                    "start": 1095.62,
                    "end": 1095.82,
                    "word": " from",
                    "probability": 0.8125
                },
                {
                    "start": 1095.92,
                    "end": 1096.38,
                    "word": " a probability",
                    "probability": 0.85888671875
                },
                {
                    "start": 1096.38,
                    "end": 1097.06,
                    "word": " distribution",
                    "probability": 0.74609375
                },
                {
                    "start": 1097.06,
                    "end": 1097.24,
                    "word": " and",
                    "probability": 0.83447265625
                },
                {
                    "start": 1097.24,
                    "end": 1097.44,
                    "word": " moving",
                    "probability": 0.86474609375
                },
                {
                    "start": 1097.44,
                    "end": 1097.62,
                    "word": " through",
                    "probability": 0.80029296875
                }
            ]
        },
        {
            "id": 180,
            "seek": 109818,
            "start": 1097.62,
            "end": 1102.58,
            "text": " this graph to generate language. You could never have generated anything close to that unicorn",
            "tokens": [
                50984,
                341,
                4295,
                281,
                8460,
                2856,
                13,
                509,
                727,
                1128,
                362,
                10833,
                1340,
                1998,
                281,
                300,
                28122,
                51236
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2641782343939499,
            "compression_ratio": 1.671280276816609,
            "no_speech_prob": 0.0002779960632324219,
            "words": [
                {
                    "start": 1097.62,
                    "end": 1097.84,
                    "word": " this",
                    "probability": 0.88037109375
                },
                {
                    "start": 1097.84,
                    "end": 1098.22,
                    "word": " graph",
                    "probability": 0.76953125
                },
                {
                    "start": 1098.22,
                    "end": 1098.72,
                    "word": " to",
                    "probability": 0.88037109375
                },
                {
                    "start": 1098.72,
                    "end": 1098.94,
                    "word": " generate",
                    "probability": 0.837890625
                },
                {
                    "start": 1098.94,
                    "end": 1099.32,
                    "word": " language",
                    "probability": 0.68701171875
                },
                {
                    "start": 1099.32,
                    "end": 1099.84,
                    "word": ".",
                    "probability": 0.93359375
                },
                {
                    "start": 1099.92,
                    "end": 1100.36,
                    "word": " You",
                    "probability": 0.88037109375
                },
                {
                    "start": 1100.36,
                    "end": 1100.48,
                    "word": " could",
                    "probability": 0.7529296875
                },
                {
                    "start": 1100.48,
                    "end": 1100.7,
                    "word": " never",
                    "probability": 0.8671875
                },
                {
                    "start": 1100.7,
                    "end": 1100.88,
                    "word": " have",
                    "probability": 0.87109375
                },
                {
                    "start": 1100.88,
                    "end": 1101.18,
                    "word": " generated",
                    "probability": 0.77880859375
                },
                {
                    "start": 1101.18,
                    "end": 1101.58,
                    "word": " anything",
                    "probability": 0.845703125
                },
                {
                    "start": 1101.58,
                    "end": 1101.98,
                    "word": " close",
                    "probability": 0.8369140625
                },
                {
                    "start": 1101.98,
                    "end": 1102.12,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 1102.12,
                    "end": 1102.28,
                    "word": " that",
                    "probability": 0.88134765625
                },
                {
                    "start": 1102.28,
                    "end": 1102.58,
                    "word": " unicorn",
                    "probability": 0.7587890625
                }
            ]
        },
        {
            "id": 181,
            "seek": 109818,
            "start": 1102.58,
            "end": 1109.12,
            "text": " story using a Markov model, and yet as recently as the last five, ten years, there's lots of",
            "tokens": [
                51236,
                1657,
                1228,
                257,
                3934,
                5179,
                2316,
                11,
                293,
                1939,
                382,
                3938,
                382,
                264,
                1036,
                1732,
                11,
                2064,
                924,
                11,
                456,
                311,
                3195,
                295,
                51556
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2641782343939499,
            "compression_ratio": 1.671280276816609,
            "no_speech_prob": 0.0002779960632324219,
            "words": [
                {
                    "start": 1102.58,
                    "end": 1102.88,
                    "word": " story",
                    "probability": 0.8056640625
                },
                {
                    "start": 1102.88,
                    "end": 1103.14,
                    "word": " using",
                    "probability": 0.82763671875
                },
                {
                    "start": 1103.28,
                    "end": 1103.54,
                    "word": " a Markov",
                    "probability": 0.92529296875
                },
                {
                    "start": 1103.54,
                    "end": 1103.86,
                    "word": " model",
                    "probability": 0.8388671875
                },
                {
                    "start": 1103.86,
                    "end": 1104.42,
                    "word": ",",
                    "probability": 0.50244140625
                },
                {
                    "start": 1104.42,
                    "end": 1104.54,
                    "word": " and",
                    "probability": 0.8779296875
                },
                {
                    "start": 1104.54,
                    "end": 1104.74,
                    "word": " yet",
                    "probability": 0.904296875
                },
                {
                    "start": 1104.74,
                    "end": 1105.2,
                    "word": " as",
                    "probability": 0.58740234375
                },
                {
                    "start": 1105.2,
                    "end": 1105.56,
                    "word": " recently",
                    "probability": 0.8037109375
                },
                {
                    "start": 1105.56,
                    "end": 1106.04,
                    "word": " as",
                    "probability": 0.90625
                },
                {
                    "start": 1106.04,
                    "end": 1106.66,
                    "word": " the",
                    "probability": 0.78076171875
                },
                {
                    "start": 1106.66,
                    "end": 1106.9,
                    "word": " last",
                    "probability": 0.8193359375
                },
                {
                    "start": 1106.9,
                    "end": 1107.42,
                    "word": " five",
                    "probability": 0.66943359375
                },
                {
                    "start": 1107.42,
                    "end": 1107.52,
                    "word": ",",
                    "probability": 0.90380859375
                },
                {
                    "start": 1107.52,
                    "end": 1107.64,
                    "word": " ten",
                    "probability": 0.228271484375
                },
                {
                    "start": 1107.64,
                    "end": 1108.0,
                    "word": " years",
                    "probability": 0.88232421875
                },
                {
                    "start": 1108.0,
                    "end": 1108.2,
                    "word": ",",
                    "probability": 0.771484375
                },
                {
                    "start": 1108.2,
                    "end": 1108.88,
                    "word": " there's",
                    "probability": 0.847412109375
                },
                {
                    "start": 1108.88,
                    "end": 1109.02,
                    "word": " lots",
                    "probability": 0.79541015625
                },
                {
                    "start": 1109.02,
                    "end": 1109.12,
                    "word": " of",
                    "probability": 0.90673828125
                }
            ]
        },
        {
            "id": 182,
            "seek": 109818,
            "start": 1109.12,
            "end": 1114.84,
            "text": " papers coming out around like using hidden Markov models to parse sentences and that kind of thing.",
            "tokens": [
                51556,
                10577,
                1348,
                484,
                926,
                411,
                1228,
                7633,
                3934,
                5179,
                5245,
                281,
                48377,
                16579,
                293,
                300,
                733,
                295,
                551,
                13,
                51832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2641782343939499,
            "compression_ratio": 1.671280276816609,
            "no_speech_prob": 0.0002779960632324219,
            "words": [
                {
                    "start": 1109.12,
                    "end": 1109.32,
                    "word": " papers",
                    "probability": 0.79541015625
                },
                {
                    "start": 1109.32,
                    "end": 1109.56,
                    "word": " coming",
                    "probability": 0.7978515625
                },
                {
                    "start": 1109.56,
                    "end": 1109.86,
                    "word": " out",
                    "probability": 0.81640625
                },
                {
                    "start": 1109.86,
                    "end": 1110.26,
                    "word": " around",
                    "probability": 0.841796875
                },
                {
                    "start": 1110.26,
                    "end": 1110.58,
                    "word": " like",
                    "probability": 0.374755859375
                },
                {
                    "start": 1110.58,
                    "end": 1110.78,
                    "word": " using",
                    "probability": 0.75927734375
                },
                {
                    "start": 1110.78,
                    "end": 1110.98,
                    "word": " hidden",
                    "probability": 0.78076171875
                },
                {
                    "start": 1110.98,
                    "end": 1111.34,
                    "word": " Markov",
                    "probability": 0.896240234375
                },
                {
                    "start": 1111.34,
                    "end": 1111.72,
                    "word": " models",
                    "probability": 0.81689453125
                },
                {
                    "start": 1111.72,
                    "end": 1112.18,
                    "word": " to",
                    "probability": 0.89599609375
                },
                {
                    "start": 1112.18,
                    "end": 1113.34,
                    "word": " parse",
                    "probability": 0.865234375
                },
                {
                    "start": 1113.34,
                    "end": 1113.78,
                    "word": " sentences",
                    "probability": 0.857421875
                },
                {
                    "start": 1113.78,
                    "end": 1114.0,
                    "word": " and",
                    "probability": 0.85009765625
                },
                {
                    "start": 1114.0,
                    "end": 1114.12,
                    "word": " that",
                    "probability": 0.880859375
                },
                {
                    "start": 1114.12,
                    "end": 1114.28,
                    "word": " kind",
                    "probability": 0.7998046875
                },
                {
                    "start": 1114.28,
                    "end": 1114.38,
                    "word": " of",
                    "probability": 0.90966796875
                },
                {
                    "start": 1114.38,
                    "end": 1114.62,
                    "word": " thing",
                    "probability": 0.81298828125
                },
                {
                    "start": 1114.62,
                    "end": 1114.84,
                    "word": ".",
                    "probability": 0.853515625
                }
            ]
        },
        {
            "id": 183,
            "seek": 112242,
            "start": 1115.16,
            "end": 1119.06,
            "text": " These are still useful models, but we've gone far beyond the simplifying assumptions",
            "tokens": [
                50364,
                1981,
                366,
                920,
                4420,
                5245,
                11,
                457,
                321,
                600,
                2780,
                1400,
                4399,
                264,
                6883,
                5489,
                17695,
                50584
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19204905251913432,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 1.3828277587890625e-05,
            "words": [
                {
                    "start": 1115.16,
                    "end": 1115.38,
                    "word": " These",
                    "probability": 0.4873046875
                },
                {
                    "start": 1115.38,
                    "end": 1115.46,
                    "word": " are",
                    "probability": 0.8974609375
                },
                {
                    "start": 1115.46,
                    "end": 1115.62,
                    "word": " still",
                    "probability": 0.88916015625
                },
                {
                    "start": 1115.62,
                    "end": 1115.86,
                    "word": " useful",
                    "probability": 0.7841796875
                },
                {
                    "start": 1115.86,
                    "end": 1116.26,
                    "word": " models",
                    "probability": 0.80322265625
                },
                {
                    "start": 1116.26,
                    "end": 1117.08,
                    "word": ",",
                    "probability": 0.59765625
                },
                {
                    "start": 1117.08,
                    "end": 1117.24,
                    "word": " but",
                    "probability": 0.86181640625
                },
                {
                    "start": 1117.24,
                    "end": 1117.46,
                    "word": " we've",
                    "probability": 0.92578125
                },
                {
                    "start": 1117.46,
                    "end": 1117.6,
                    "word": " gone",
                    "probability": 0.875
                },
                {
                    "start": 1117.6,
                    "end": 1117.84,
                    "word": " far",
                    "probability": 0.88916015625
                },
                {
                    "start": 1117.84,
                    "end": 1118.12,
                    "word": " beyond",
                    "probability": 0.75
                },
                {
                    "start": 1118.12,
                    "end": 1118.3,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 1118.3,
                    "end": 1118.64,
                    "word": " simplifying",
                    "probability": 0.913330078125
                },
                {
                    "start": 1118.64,
                    "end": 1119.06,
                    "word": " assumptions",
                    "probability": 0.91552734375
                }
            ]
        },
        {
            "id": 184,
            "seek": 112242,
            "start": 1119.52,
            "end": 1127.54,
            "text": " in these original sort of simplified models of the world in NLP from the past few decades.",
            "tokens": [
                50584,
                294,
                613,
                3380,
                1333,
                295,
                26335,
                5245,
                295,
                264,
                1002,
                294,
                426,
                45196,
                490,
                264,
                1791,
                1326,
                7878,
                13,
                50960
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19204905251913432,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 1.3828277587890625e-05,
            "words": [
                {
                    "start": 1119.52,
                    "end": 1119.76,
                    "word": " in",
                    "probability": 0.82958984375
                },
                {
                    "start": 1119.76,
                    "end": 1119.92,
                    "word": " these",
                    "probability": 0.70947265625
                },
                {
                    "start": 1119.92,
                    "end": 1120.46,
                    "word": " original",
                    "probability": 0.8134765625
                },
                {
                    "start": 1120.46,
                    "end": 1122.72,
                    "word": " sort",
                    "probability": 0.439208984375
                },
                {
                    "start": 1122.72,
                    "end": 1122.82,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 1122.82,
                    "end": 1123.14,
                    "word": " simplified",
                    "probability": 0.8798828125
                },
                {
                    "start": 1123.14,
                    "end": 1123.54,
                    "word": " models",
                    "probability": 0.8056640625
                },
                {
                    "start": 1123.54,
                    "end": 1123.68,
                    "word": " of",
                    "probability": 0.8857421875
                },
                {
                    "start": 1123.68,
                    "end": 1123.78,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 1123.78,
                    "end": 1124.02,
                    "word": " world",
                    "probability": 0.63720703125
                },
                {
                    "start": 1124.02,
                    "end": 1125.0,
                    "word": " in",
                    "probability": 0.58544921875
                },
                {
                    "start": 1125.0,
                    "end": 1125.98,
                    "word": " NLP",
                    "probability": 0.881103515625
                },
                {
                    "start": 1125.98,
                    "end": 1126.22,
                    "word": " from",
                    "probability": 0.79638671875
                },
                {
                    "start": 1126.22,
                    "end": 1126.36,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 1126.36,
                    "end": 1126.54,
                    "word": " past",
                    "probability": 0.82666015625
                },
                {
                    "start": 1126.54,
                    "end": 1126.7,
                    "word": " few",
                    "probability": 0.79736328125
                },
                {
                    "start": 1126.7,
                    "end": 1127.06,
                    "word": " decades",
                    "probability": 0.91748046875
                },
                {
                    "start": 1127.06,
                    "end": 1127.54,
                    "word": ".",
                    "probability": 0.9951171875
                }
            ]
        },
        {
            "id": 185,
            "seek": 112242,
            "start": 1129.5,
            "end": 1134.82,
            "text": " So let's contrast transformers now, and I'll get into more details about how transformers",
            "tokens": [
                51092,
                407,
                718,
                311,
                8712,
                4088,
                433,
                586,
                11,
                293,
                286,
                603,
                483,
                666,
                544,
                4365,
                466,
                577,
                4088,
                433,
                51344
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19204905251913432,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 1.3828277587890625e-05,
            "words": [
                {
                    "start": 1129.5,
                    "end": 1129.98,
                    "word": " So",
                    "probability": 0.853515625
                },
                {
                    "start": 1129.98,
                    "end": 1130.3,
                    "word": " let's",
                    "probability": 0.873291015625
                },
                {
                    "start": 1130.3,
                    "end": 1130.72,
                    "word": " contrast",
                    "probability": 0.73828125
                },
                {
                    "start": 1130.72,
                    "end": 1132.54,
                    "word": " transformers",
                    "probability": 0.784912109375
                },
                {
                    "start": 1132.54,
                    "end": 1132.8,
                    "word": " now",
                    "probability": 0.89013671875
                },
                {
                    "start": 1132.8,
                    "end": 1132.96,
                    "word": ",",
                    "probability": 0.55126953125
                },
                {
                    "start": 1132.96,
                    "end": 1133.04,
                    "word": " and",
                    "probability": 0.8955078125
                },
                {
                    "start": 1133.04,
                    "end": 1133.14,
                    "word": " I'll",
                    "probability": 0.973388671875
                },
                {
                    "start": 1133.14,
                    "end": 1133.3,
                    "word": " get",
                    "probability": 0.89794921875
                },
                {
                    "start": 1133.3,
                    "end": 1133.44,
                    "word": " into",
                    "probability": 0.73486328125
                },
                {
                    "start": 1133.44,
                    "end": 1133.64,
                    "word": " more",
                    "probability": 0.8759765625
                },
                {
                    "start": 1133.64,
                    "end": 1133.96,
                    "word": " details",
                    "probability": 0.7353515625
                },
                {
                    "start": 1133.96,
                    "end": 1134.18,
                    "word": " about",
                    "probability": 0.8203125
                },
                {
                    "start": 1134.18,
                    "end": 1134.36,
                    "word": " how",
                    "probability": 0.892578125
                },
                {
                    "start": 1134.36,
                    "end": 1134.82,
                    "word": " transformers",
                    "probability": 0.903564453125
                }
            ]
        },
        {
            "id": 186,
            "seek": 112242,
            "start": 1134.82,
            "end": 1139.4,
            "text": " work in a second, but let's contrast transformers with these earlier natural",
            "tokens": [
                51344,
                589,
                294,
                257,
                1150,
                11,
                457,
                718,
                311,
                8712,
                4088,
                433,
                365,
                613,
                3071,
                3303,
                51572
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19204905251913432,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 1.3828277587890625e-05,
            "words": [
                {
                    "start": 1134.82,
                    "end": 1135.04,
                    "word": " work",
                    "probability": 0.79248046875
                },
                {
                    "start": 1135.04,
                    "end": 1135.16,
                    "word": " in",
                    "probability": 0.90673828125
                },
                {
                    "start": 1135.26,
                    "end": 1135.5,
                    "word": " a second",
                    "probability": 0.80224609375
                },
                {
                    "start": 1135.5,
                    "end": 1135.64,
                    "word": ",",
                    "probability": 0.62646484375
                },
                {
                    "start": 1135.64,
                    "end": 1135.72,
                    "word": " but",
                    "probability": 0.8896484375
                },
                {
                    "start": 1135.72,
                    "end": 1136.86,
                    "word": " let's",
                    "probability": 0.95068359375
                },
                {
                    "start": 1136.86,
                    "end": 1137.12,
                    "word": " contrast",
                    "probability": 0.73876953125
                },
                {
                    "start": 1137.12,
                    "end": 1137.78,
                    "word": " transformers",
                    "probability": 0.9306640625
                },
                {
                    "start": 1137.78,
                    "end": 1137.98,
                    "word": " with",
                    "probability": 0.81591796875
                },
                {
                    "start": 1137.98,
                    "end": 1138.32,
                    "word": " these",
                    "probability": 0.70556640625
                },
                {
                    "start": 1138.32,
                    "end": 1139.1,
                    "word": " earlier",
                    "probability": 0.90185546875
                },
                {
                    "start": 1139.1,
                    "end": 1139.4,
                    "word": " natural",
                    "probability": 0.8388671875
                }
            ]
        },
        {
            "id": 187,
            "seek": 114600,
            "start": 1139.4,
            "end": 1146.32,
            "text": " language processing approaches. So pre-transformer, most machine learning",
            "tokens": [
                50368,
                2856,
                9007,
                11587,
                13,
                407,
                659,
                12,
                24999,
                837,
                260,
                11,
                881,
                3479,
                2539,
                50708
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2423270046710968,
            "compression_ratio": 1.526946107784431,
            "no_speech_prob": 0.0010986328125,
            "words": [
                {
                    "start": 1139.4,
                    "end": 1139.58,
                    "word": " language",
                    "probability": 0.399658203125
                },
                {
                    "start": 1139.58,
                    "end": 1139.94,
                    "word": " processing",
                    "probability": 0.8017578125
                },
                {
                    "start": 1139.94,
                    "end": 1140.34,
                    "word": " approaches",
                    "probability": 0.69091796875
                },
                {
                    "start": 1140.34,
                    "end": 1141.06,
                    "word": ".",
                    "probability": 0.89794921875
                },
                {
                    "start": 1141.1,
                    "end": 1141.24,
                    "word": " So",
                    "probability": 0.857421875
                },
                {
                    "start": 1141.24,
                    "end": 1142.34,
                    "word": " pre",
                    "probability": 0.65673828125
                },
                {
                    "start": 1142.34,
                    "end": 1143.76,
                    "word": "-transformer",
                    "probability": 0.7987060546875
                },
                {
                    "start": 1143.76,
                    "end": 1144.06,
                    "word": ",",
                    "probability": 0.7333984375
                },
                {
                    "start": 1144.06,
                    "end": 1144.74,
                    "word": " most",
                    "probability": 0.802734375
                },
                {
                    "start": 1144.74,
                    "end": 1146.06,
                    "word": " machine",
                    "probability": 0.71875
                },
                {
                    "start": 1146.06,
                    "end": 1146.32,
                    "word": " learning",
                    "probability": 0.8212890625
                }
            ]
        },
        {
            "id": 188,
            "seek": 114600,
            "start": 1146.32,
            "end": 1154.62,
            "text": " approaches didn't consider words in context. Many approaches made the simplifying bag of",
            "tokens": [
                50708,
                11587,
                994,
                380,
                1949,
                2283,
                294,
                4319,
                13,
                5126,
                11587,
                1027,
                264,
                6883,
                5489,
                3411,
                295,
                51140
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2423270046710968,
            "compression_ratio": 1.526946107784431,
            "no_speech_prob": 0.0010986328125,
            "words": [
                {
                    "start": 1146.32,
                    "end": 1146.7,
                    "word": " approaches",
                    "probability": 0.66552734375
                },
                {
                    "start": 1146.7,
                    "end": 1147.08,
                    "word": " didn't",
                    "probability": 0.9130859375
                },
                {
                    "start": 1147.08,
                    "end": 1147.38,
                    "word": " consider",
                    "probability": 0.8212890625
                },
                {
                    "start": 1147.38,
                    "end": 1147.72,
                    "word": " words",
                    "probability": 0.7392578125
                },
                {
                    "start": 1147.72,
                    "end": 1147.86,
                    "word": " in",
                    "probability": 0.412841796875
                },
                {
                    "start": 1147.86,
                    "end": 1148.42,
                    "word": " context",
                    "probability": 0.76611328125
                },
                {
                    "start": 1148.42,
                    "end": 1149.04,
                    "word": ".",
                    "probability": 0.880859375
                },
                {
                    "start": 1149.26,
                    "end": 1149.98,
                    "word": " Many",
                    "probability": 0.87646484375
                },
                {
                    "start": 1149.98,
                    "end": 1150.64,
                    "word": " approaches",
                    "probability": 0.67236328125
                },
                {
                    "start": 1150.64,
                    "end": 1152.96,
                    "word": " made",
                    "probability": 0.791015625
                },
                {
                    "start": 1152.96,
                    "end": 1153.64,
                    "word": " the",
                    "probability": 0.79833984375
                },
                {
                    "start": 1153.64,
                    "end": 1154.2,
                    "word": " simplifying",
                    "probability": 0.939697265625
                },
                {
                    "start": 1154.2,
                    "end": 1154.48,
                    "word": " bag",
                    "probability": 0.84228515625
                },
                {
                    "start": 1154.48,
                    "end": 1154.62,
                    "word": " of",
                    "probability": 0.7744140625
                }
            ]
        },
        {
            "id": 189,
            "seek": 114600,
            "start": 1154.62,
            "end": 1162.98,
            "text": " words assumptions as kind of a first step in the modeling process, and then ran term vectors",
            "tokens": [
                51140,
                2283,
                17695,
                382,
                733,
                295,
                257,
                700,
                1823,
                294,
                264,
                15983,
                1399,
                11,
                293,
                550,
                5872,
                1433,
                18875,
                51548
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2423270046710968,
            "compression_ratio": 1.526946107784431,
            "no_speech_prob": 0.0010986328125,
            "words": [
                {
                    "start": 1154.62,
                    "end": 1154.8,
                    "word": " words",
                    "probability": 0.509765625
                },
                {
                    "start": 1154.8,
                    "end": 1155.64,
                    "word": " assumptions",
                    "probability": 0.6904296875
                },
                {
                    "start": 1155.64,
                    "end": 1157.84,
                    "word": " as",
                    "probability": 0.85693359375
                },
                {
                    "start": 1157.84,
                    "end": 1158.02,
                    "word": " kind",
                    "probability": 0.69140625
                },
                {
                    "start": 1158.02,
                    "end": 1158.08,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1158.18,
                    "end": 1158.42,
                    "word": " a first",
                    "probability": 0.8115234375
                },
                {
                    "start": 1158.42,
                    "end": 1158.78,
                    "word": " step",
                    "probability": 0.81689453125
                },
                {
                    "start": 1158.78,
                    "end": 1159.22,
                    "word": " in",
                    "probability": 0.89990234375
                },
                {
                    "start": 1159.22,
                    "end": 1159.58,
                    "word": " the",
                    "probability": 0.8046875
                },
                {
                    "start": 1159.58,
                    "end": 1159.82,
                    "word": " modeling",
                    "probability": 0.68212890625
                },
                {
                    "start": 1159.82,
                    "end": 1160.66,
                    "word": " process",
                    "probability": 0.87109375
                },
                {
                    "start": 1160.66,
                    "end": 1160.88,
                    "word": ",",
                    "probability": 0.3203125
                },
                {
                    "start": 1160.88,
                    "end": 1161.04,
                    "word": " and",
                    "probability": 0.89306640625
                },
                {
                    "start": 1161.04,
                    "end": 1161.18,
                    "word": " then",
                    "probability": 0.70947265625
                },
                {
                    "start": 1161.18,
                    "end": 1161.72,
                    "word": " ran",
                    "probability": 0.8603515625
                },
                {
                    "start": 1161.72,
                    "end": 1162.64,
                    "word": " term",
                    "probability": 0.6787109375
                },
                {
                    "start": 1162.64,
                    "end": 1162.98,
                    "word": " vectors",
                    "probability": 0.86767578125
                }
            ]
        },
        {
            "id": 190,
            "seek": 117600,
            "start": 1162.98,
            "end": 1169.98,
            "text": " through models like topic models, or logistic regression, or support vector machines.",
            "tokens": [
                50368,
                807,
                5245,
                411,
                4829,
                5245,
                11,
                420,
                3565,
                3142,
                24590,
                11,
                420,
                1406,
                8062,
                8379,
                13,
                50688
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21894531678408385,
            "compression_ratio": 1.671497584541063,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1162.98,
                    "end": 1163.34,
                    "word": " through",
                    "probability": 0.35693359375
                },
                {
                    "start": 1163.34,
                    "end": 1164.18,
                    "word": " models",
                    "probability": 0.71044921875
                },
                {
                    "start": 1164.18,
                    "end": 1164.56,
                    "word": " like",
                    "probability": 0.85498046875
                },
                {
                    "start": 1164.56,
                    "end": 1165.88,
                    "word": " topic",
                    "probability": 0.6376953125
                },
                {
                    "start": 1165.88,
                    "end": 1166.32,
                    "word": " models",
                    "probability": 0.80224609375
                },
                {
                    "start": 1166.32,
                    "end": 1166.6,
                    "word": ",",
                    "probability": 0.2325439453125
                },
                {
                    "start": 1166.6,
                    "end": 1166.72,
                    "word": " or",
                    "probability": 0.9052734375
                },
                {
                    "start": 1166.72,
                    "end": 1167.06,
                    "word": " logistic",
                    "probability": 0.95068359375
                },
                {
                    "start": 1167.06,
                    "end": 1167.48,
                    "word": " regression",
                    "probability": 0.85595703125
                },
                {
                    "start": 1167.48,
                    "end": 1167.9,
                    "word": ",",
                    "probability": 0.9091796875
                },
                {
                    "start": 1167.9,
                    "end": 1168.48,
                    "word": " or",
                    "probability": 0.896484375
                },
                {
                    "start": 1168.48,
                    "end": 1168.76,
                    "word": " support",
                    "probability": 0.8662109375
                },
                {
                    "start": 1168.76,
                    "end": 1169.04,
                    "word": " vector",
                    "probability": 0.89794921875
                },
                {
                    "start": 1169.04,
                    "end": 1169.48,
                    "word": " machines",
                    "probability": 0.8212890625
                },
                {
                    "start": 1169.48,
                    "end": 1169.98,
                    "word": ".",
                    "probability": 0.9833984375
                }
            ]
        },
        {
            "id": 191,
            "seek": 117600,
            "start": 1170.62,
            "end": 1179.24,
            "text": " Most models didn't model kind of like co-reference relationships between words,",
            "tokens": [
                50768,
                4534,
                5245,
                994,
                380,
                2316,
                733,
                295,
                411,
                598,
                12,
                265,
                5158,
                6159,
                1296,
                2283,
                11,
                51168
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21894531678408385,
            "compression_ratio": 1.671497584541063,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1170.62,
                    "end": 1171.38,
                    "word": " Most",
                    "probability": 0.83544921875
                },
                {
                    "start": 1171.38,
                    "end": 1171.78,
                    "word": " models",
                    "probability": 0.802734375
                },
                {
                    "start": 1171.78,
                    "end": 1172.46,
                    "word": " didn't",
                    "probability": 0.925537109375
                },
                {
                    "start": 1172.46,
                    "end": 1173.82,
                    "word": " model",
                    "probability": 0.8515625
                },
                {
                    "start": 1173.82,
                    "end": 1174.28,
                    "word": " kind",
                    "probability": 0.369873046875
                },
                {
                    "start": 1174.28,
                    "end": 1174.56,
                    "word": " of",
                    "probability": 0.91064453125
                },
                {
                    "start": 1174.56,
                    "end": 1175.0,
                    "word": " like",
                    "probability": 0.7763671875
                },
                {
                    "start": 1175.0,
                    "end": 1175.3,
                    "word": " co",
                    "probability": 0.61376953125
                },
                {
                    "start": 1175.3,
                    "end": 1175.82,
                    "word": "-reference",
                    "probability": 0.8709309895833334
                },
                {
                    "start": 1175.82,
                    "end": 1178.06,
                    "word": " relationships",
                    "probability": 0.6552734375
                },
                {
                    "start": 1178.06,
                    "end": 1178.48,
                    "word": " between",
                    "probability": 0.78173828125
                },
                {
                    "start": 1178.48,
                    "end": 1179.0,
                    "word": " words",
                    "probability": 0.724609375
                },
                {
                    "start": 1179.0,
                    "end": 1179.24,
                    "word": ",",
                    "probability": 0.485595703125
                }
            ]
        },
        {
            "id": 192,
            "seek": 117600,
            "start": 1179.24,
            "end": 1184.52,
            "text": " or sort of which words pertain to which other words in a sentence, and I'll talk about what",
            "tokens": [
                51168,
                420,
                1333,
                295,
                597,
                2283,
                13269,
                491,
                281,
                597,
                661,
                2283,
                294,
                257,
                8174,
                11,
                293,
                286,
                603,
                751,
                466,
                437,
                51440
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21894531678408385,
            "compression_ratio": 1.671497584541063,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1179.24,
                    "end": 1179.42,
                    "word": " or",
                    "probability": 0.8984375
                },
                {
                    "start": 1179.42,
                    "end": 1179.82,
                    "word": " sort",
                    "probability": 0.74072265625
                },
                {
                    "start": 1179.82,
                    "end": 1179.92,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1179.92,
                    "end": 1180.16,
                    "word": " which",
                    "probability": 0.86376953125
                },
                {
                    "start": 1180.16,
                    "end": 1180.58,
                    "word": " words",
                    "probability": 0.71826171875
                },
                {
                    "start": 1180.58,
                    "end": 1180.96,
                    "word": " pertain",
                    "probability": 0.8974609375
                },
                {
                    "start": 1180.96,
                    "end": 1181.12,
                    "word": " to",
                    "probability": 0.9052734375
                },
                {
                    "start": 1181.12,
                    "end": 1181.26,
                    "word": " which",
                    "probability": 0.89501953125
                },
                {
                    "start": 1181.26,
                    "end": 1181.5,
                    "word": " other",
                    "probability": 0.78271484375
                },
                {
                    "start": 1181.5,
                    "end": 1181.9,
                    "word": " words",
                    "probability": 0.7353515625
                },
                {
                    "start": 1181.9,
                    "end": 1182.18,
                    "word": " in",
                    "probability": 0.88671875
                },
                {
                    "start": 1182.34,
                    "end": 1182.8,
                    "word": " a sentence",
                    "probability": 0.8330078125
                },
                {
                    "start": 1182.8,
                    "end": 1183.18,
                    "word": ",",
                    "probability": 0.412109375
                },
                {
                    "start": 1183.18,
                    "end": 1183.8,
                    "word": " and",
                    "probability": 0.89697265625
                },
                {
                    "start": 1183.8,
                    "end": 1183.9,
                    "word": " I'll",
                    "probability": 0.972412109375
                },
                {
                    "start": 1183.9,
                    "end": 1184.08,
                    "word": " talk",
                    "probability": 0.80908203125
                },
                {
                    "start": 1184.08,
                    "end": 1184.28,
                    "word": " about",
                    "probability": 0.826171875
                },
                {
                    "start": 1184.28,
                    "end": 1184.52,
                    "word": " what",
                    "probability": 0.7880859375
                }
            ]
        },
        {
            "id": 193,
            "seek": 117600,
            "start": 1184.52,
            "end": 1191.14,
            "text": " that means later, but transformers do sort of solve that problem. Most approaches didn't",
            "tokens": [
                51440,
                300,
                1355,
                1780,
                11,
                457,
                4088,
                433,
                360,
                1333,
                295,
                5039,
                300,
                1154,
                13,
                4534,
                11587,
                994,
                380,
                51772
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21894531678408385,
            "compression_ratio": 1.671497584541063,
            "no_speech_prob": 0.0007319450378417969,
            "words": [
                {
                    "start": 1184.52,
                    "end": 1184.7,
                    "word": " that",
                    "probability": 0.880859375
                },
                {
                    "start": 1184.7,
                    "end": 1184.88,
                    "word": " means",
                    "probability": 0.88232421875
                },
                {
                    "start": 1184.88,
                    "end": 1185.08,
                    "word": " later",
                    "probability": 0.87158203125
                },
                {
                    "start": 1185.08,
                    "end": 1185.14,
                    "word": ",",
                    "probability": 0.8056640625
                },
                {
                    "start": 1185.14,
                    "end": 1185.24,
                    "word": " but",
                    "probability": 0.89111328125
                },
                {
                    "start": 1185.24,
                    "end": 1185.7,
                    "word": " transformers",
                    "probability": 0.830322265625
                },
                {
                    "start": 1185.7,
                    "end": 1185.96,
                    "word": " do",
                    "probability": 0.90087890625
                },
                {
                    "start": 1185.96,
                    "end": 1186.4,
                    "word": " sort",
                    "probability": 0.79052734375
                },
                {
                    "start": 1186.4,
                    "end": 1186.48,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 1186.48,
                    "end": 1186.64,
                    "word": " solve",
                    "probability": 0.84521484375
                },
                {
                    "start": 1186.64,
                    "end": 1186.8,
                    "word": " that",
                    "probability": 0.88232421875
                },
                {
                    "start": 1186.8,
                    "end": 1187.12,
                    "word": " problem",
                    "probability": 0.76904296875
                },
                {
                    "start": 1187.12,
                    "end": 1187.88,
                    "word": ".",
                    "probability": 0.99267578125
                },
                {
                    "start": 1189.06,
                    "end": 1189.78,
                    "word": " Most",
                    "probability": 0.8427734375
                },
                {
                    "start": 1189.78,
                    "end": 1190.5,
                    "word": " approaches",
                    "probability": 0.6064453125
                },
                {
                    "start": 1190.5,
                    "end": 1191.14,
                    "word": " didn't",
                    "probability": 0.919677734375
                }
            ]
        },
        {
            "id": 194,
            "seek": 120522,
            "start": 1193.32,
            "end": 1200.34,
            "text": " operate on either words or characters. Typically the way people use transformers,",
            "tokens": [
                50404,
                9651,
                322,
                2139,
                2283,
                420,
                4342,
                13,
                23129,
                264,
                636,
                561,
                764,
                4088,
                433,
                11,
                50716
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2473958202061199,
            "compression_ratio": 1.5982905982905984,
            "no_speech_prob": 4.398822784423828e-05,
            "words": [
                {
                    "start": 1193.32,
                    "end": 1193.84,
                    "word": " operate",
                    "probability": 1.9431114196777344e-05
                },
                {
                    "start": 1193.84,
                    "end": 1194.36,
                    "word": " on",
                    "probability": 0.1507568359375
                },
                {
                    "start": 1194.36,
                    "end": 1194.58,
                    "word": " either",
                    "probability": 0.2744140625
                },
                {
                    "start": 1194.58,
                    "end": 1194.98,
                    "word": " words",
                    "probability": 0.69482421875
                },
                {
                    "start": 1194.98,
                    "end": 1195.2,
                    "word": " or",
                    "probability": 0.88232421875
                },
                {
                    "start": 1195.2,
                    "end": 1195.9,
                    "word": " characters",
                    "probability": 0.82373046875
                },
                {
                    "start": 1195.9,
                    "end": 1196.42,
                    "word": ".",
                    "probability": 0.85009765625
                },
                {
                    "start": 1197.12,
                    "end": 1197.58,
                    "word": " Typically",
                    "probability": 0.07232666015625
                },
                {
                    "start": 1197.58,
                    "end": 1199.02,
                    "word": " the",
                    "probability": 0.57275390625
                },
                {
                    "start": 1199.02,
                    "end": 1199.1,
                    "word": " way",
                    "probability": 0.89794921875
                },
                {
                    "start": 1199.1,
                    "end": 1199.26,
                    "word": " people",
                    "probability": 0.8779296875
                },
                {
                    "start": 1199.26,
                    "end": 1199.44,
                    "word": " use",
                    "probability": 0.794921875
                },
                {
                    "start": 1199.44,
                    "end": 1200.08,
                    "word": " transformers",
                    "probability": 0.85107421875
                },
                {
                    "start": 1200.08,
                    "end": 1200.34,
                    "word": ",",
                    "probability": 0.595703125
                }
            ]
        },
        {
            "id": 195,
            "seek": 120522,
            "start": 1200.34,
            "end": 1206.6,
            "text": " we use well-chosen chunks of words, which allows us to model misspelling and this kind of thing,",
            "tokens": [
                50748,
                321,
                764,
                731,
                12,
                339,
                6441,
                24004,
                295,
                2283,
                11,
                597,
                4045,
                505,
                281,
                2316,
                1713,
                494,
                2669,
                293,
                341,
                733,
                295,
                551,
                11,
                51044
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2473958202061199,
            "compression_ratio": 1.5982905982905984,
            "no_speech_prob": 4.398822784423828e-05,
            "words": [
                {
                    "start": 1200.34,
                    "end": 1200.88,
                    "word": " we",
                    "probability": 0.88623046875
                },
                {
                    "start": 1200.88,
                    "end": 1201.26,
                    "word": " use",
                    "probability": 0.826171875
                },
                {
                    "start": 1201.26,
                    "end": 1202.28,
                    "word": " well",
                    "probability": 0.83203125
                },
                {
                    "start": 1202.28,
                    "end": 1202.64,
                    "word": "-chosen",
                    "probability": 0.7696940104166666
                },
                {
                    "start": 1202.64,
                    "end": 1203.0,
                    "word": " chunks",
                    "probability": 0.83642578125
                },
                {
                    "start": 1203.0,
                    "end": 1203.2,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 1203.2,
                    "end": 1203.54,
                    "word": " words",
                    "probability": 0.720703125
                },
                {
                    "start": 1203.54,
                    "end": 1203.76,
                    "word": ",",
                    "probability": 0.57763671875
                },
                {
                    "start": 1203.76,
                    "end": 1204.22,
                    "word": " which",
                    "probability": 0.91162109375
                },
                {
                    "start": 1204.22,
                    "end": 1204.48,
                    "word": " allows",
                    "probability": 0.75048828125
                },
                {
                    "start": 1204.48,
                    "end": 1204.64,
                    "word": " us",
                    "probability": 0.884765625
                },
                {
                    "start": 1204.64,
                    "end": 1204.76,
                    "word": " to",
                    "probability": 0.9072265625
                },
                {
                    "start": 1204.76,
                    "end": 1204.96,
                    "word": " model",
                    "probability": 0.8017578125
                },
                {
                    "start": 1204.96,
                    "end": 1205.5,
                    "word": " misspelling",
                    "probability": 0.8603515625
                },
                {
                    "start": 1205.5,
                    "end": 1205.64,
                    "word": " and",
                    "probability": 0.425537109375
                },
                {
                    "start": 1205.64,
                    "end": 1205.8,
                    "word": " this",
                    "probability": 0.74169921875
                },
                {
                    "start": 1205.8,
                    "end": 1205.96,
                    "word": " kind",
                    "probability": 0.78955078125
                },
                {
                    "start": 1205.96,
                    "end": 1206.06,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 1206.06,
                    "end": 1206.3,
                    "word": " thing",
                    "probability": 0.8203125
                },
                {
                    "start": 1206.3,
                    "end": 1206.6,
                    "word": ",",
                    "probability": 0.30029296875
                }
            ]
        },
        {
            "id": 196,
            "seek": 120522,
            "start": 1206.6,
            "end": 1210.96,
            "text": " so there's been an improvement there in the current generation of natural language processing models.",
            "tokens": [
                51044,
                370,
                456,
                311,
                668,
                364,
                10444,
                456,
                294,
                264,
                2190,
                5125,
                295,
                3303,
                2856,
                9007,
                5245,
                13,
                51236
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2473958202061199,
            "compression_ratio": 1.5982905982905984,
            "no_speech_prob": 4.398822784423828e-05,
            "words": [
                {
                    "start": 1206.6,
                    "end": 1206.88,
                    "word": " so",
                    "probability": 0.84423828125
                },
                {
                    "start": 1206.88,
                    "end": 1207.04,
                    "word": " there's",
                    "probability": 0.853271484375
                },
                {
                    "start": 1207.04,
                    "end": 1207.16,
                    "word": " been",
                    "probability": 0.796875
                },
                {
                    "start": 1207.16,
                    "end": 1207.3,
                    "word": " an",
                    "probability": 0.869140625
                },
                {
                    "start": 1207.3,
                    "end": 1207.56,
                    "word": " improvement",
                    "probability": 0.68359375
                },
                {
                    "start": 1207.56,
                    "end": 1207.9,
                    "word": " there",
                    "probability": 0.833984375
                },
                {
                    "start": 1207.9,
                    "end": 1208.54,
                    "word": " in",
                    "probability": 0.73095703125
                },
                {
                    "start": 1208.54,
                    "end": 1208.64,
                    "word": " the",
                    "probability": 0.79736328125
                },
                {
                    "start": 1208.64,
                    "end": 1208.78,
                    "word": " current",
                    "probability": 0.732421875
                },
                {
                    "start": 1208.78,
                    "end": 1209.22,
                    "word": " generation",
                    "probability": 0.76416015625
                },
                {
                    "start": 1209.22,
                    "end": 1209.34,
                    "word": " of",
                    "probability": 0.85693359375
                },
                {
                    "start": 1209.34,
                    "end": 1209.56,
                    "word": " natural",
                    "probability": 0.84375
                },
                {
                    "start": 1209.56,
                    "end": 1209.76,
                    "word": " language",
                    "probability": 0.65185546875
                },
                {
                    "start": 1209.76,
                    "end": 1210.12,
                    "word": " processing",
                    "probability": 0.75927734375
                },
                {
                    "start": 1210.12,
                    "end": 1210.44,
                    "word": " models",
                    "probability": 0.78662109375
                },
                {
                    "start": 1210.44,
                    "end": 1210.96,
                    "word": ".",
                    "probability": 0.857421875
                }
            ]
        },
        {
            "id": 197,
            "seek": 120522,
            "start": 1212.06,
            "end": 1222.2,
            "text": " And older approaches tended not to use neural network technology. That's changed a lot in the",
            "tokens": [
                51292,
                400,
                4906,
                11587,
                34732,
                406,
                281,
                764,
                18161,
                3209,
                2899,
                13,
                663,
                311,
                3105,
                257,
                688,
                294,
                264,
                51524
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2473958202061199,
            "compression_ratio": 1.5982905982905984,
            "no_speech_prob": 4.398822784423828e-05,
            "words": [
                {
                    "start": 1212.06,
                    "end": 1212.58,
                    "word": " And",
                    "probability": 0.56494140625
                },
                {
                    "start": 1212.58,
                    "end": 1213.1,
                    "word": " older",
                    "probability": 0.82275390625
                },
                {
                    "start": 1213.1,
                    "end": 1213.42,
                    "word": " approaches",
                    "probability": 0.6689453125
                },
                {
                    "start": 1213.42,
                    "end": 1213.72,
                    "word": " tended",
                    "probability": 0.68017578125
                },
                {
                    "start": 1213.72,
                    "end": 1213.92,
                    "word": " not",
                    "probability": 0.8740234375
                },
                {
                    "start": 1213.92,
                    "end": 1214.06,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 1214.06,
                    "end": 1214.3,
                    "word": " use",
                    "probability": 0.828125
                },
                {
                    "start": 1214.3,
                    "end": 1214.54,
                    "word": " neural",
                    "probability": 0.84619140625
                },
                {
                    "start": 1214.54,
                    "end": 1214.78,
                    "word": " network",
                    "probability": 0.8291015625
                },
                {
                    "start": 1214.78,
                    "end": 1215.3,
                    "word": " technology",
                    "probability": 0.89990234375
                },
                {
                    "start": 1215.3,
                    "end": 1215.58,
                    "word": ".",
                    "probability": 0.62939453125
                },
                {
                    "start": 1215.58,
                    "end": 1215.76,
                    "word": " That's",
                    "probability": 0.890869140625
                },
                {
                    "start": 1215.76,
                    "end": 1215.94,
                    "word": " changed",
                    "probability": 0.80615234375
                },
                {
                    "start": 1216.06,
                    "end": 1216.14,
                    "word": " a lot",
                    "probability": 0.89404296875
                },
                {
                    "start": 1216.14,
                    "end": 1216.24,
                    "word": " in",
                    "probability": 0.888671875
                },
                {
                    "start": 1216.24,
                    "end": 1222.2,
                    "word": " the",
                    "probability": 0.8134765625
                }
            ]
        },
        {
            "id": 198,
            "seek": 123328,
            "start": 1222.2,
            "end": 1229.4,
            "text": " best ideas that have come out of the neural network revolution that's been ongoing since",
            "tokens": [
                50384,
                1151,
                3487,
                300,
                362,
                808,
                484,
                295,
                264,
                18161,
                3209,
                8894,
                300,
                311,
                668,
                10452,
                1670,
                50724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22553454496358571,
            "compression_ratio": 1.611353711790393,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1222.2,
                    "end": 1222.32,
                    "word": " best",
                    "probability": 0.06402587890625
                },
                {
                    "start": 1222.32,
                    "end": 1223.42,
                    "word": " ideas",
                    "probability": 0.73486328125
                },
                {
                    "start": 1223.42,
                    "end": 1224.1,
                    "word": " that",
                    "probability": 0.59033203125
                },
                {
                    "start": 1224.1,
                    "end": 1225.52,
                    "word": " have",
                    "probability": 0.8662109375
                },
                {
                    "start": 1225.52,
                    "end": 1225.66,
                    "word": " come",
                    "probability": 0.8037109375
                },
                {
                    "start": 1225.66,
                    "end": 1225.96,
                    "word": " out",
                    "probability": 0.82177734375
                },
                {
                    "start": 1225.96,
                    "end": 1226.58,
                    "word": " of",
                    "probability": 0.88232421875
                },
                {
                    "start": 1226.58,
                    "end": 1226.68,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1226.68,
                    "end": 1226.8,
                    "word": " neural",
                    "probability": 0.7841796875
                },
                {
                    "start": 1226.8,
                    "end": 1227.04,
                    "word": " network",
                    "probability": 0.81640625
                },
                {
                    "start": 1227.04,
                    "end": 1227.6,
                    "word": " revolution",
                    "probability": 0.81640625
                },
                {
                    "start": 1227.6,
                    "end": 1228.62,
                    "word": " that's",
                    "probability": 0.872802734375
                },
                {
                    "start": 1228.62,
                    "end": 1228.76,
                    "word": " been",
                    "probability": 0.82666015625
                },
                {
                    "start": 1228.76,
                    "end": 1229.1,
                    "word": " ongoing",
                    "probability": 0.7099609375
                },
                {
                    "start": 1229.1,
                    "end": 1229.4,
                    "word": " since",
                    "probability": 0.806640625
                }
            ]
        },
        {
            "id": 199,
            "seek": 123328,
            "start": 1229.4,
            "end": 1238.98,
            "text": " around 2012. So transformers kind of kick apart a number of log jams in natural language modeling.",
            "tokens": [
                50724,
                926,
                9125,
                13,
                407,
                4088,
                433,
                733,
                295,
                4437,
                4936,
                257,
                1230,
                295,
                3565,
                361,
                4070,
                294,
                3303,
                2856,
                15983,
                13,
                51192
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22553454496358571,
            "compression_ratio": 1.611353711790393,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1229.4,
                    "end": 1229.72,
                    "word": " around",
                    "probability": 0.83203125
                },
                {
                    "start": 1229.72,
                    "end": 1230.66,
                    "word": " 2012",
                    "probability": 0.76806640625
                },
                {
                    "start": 1230.66,
                    "end": 1231.28,
                    "word": ".",
                    "probability": 0.95068359375
                },
                {
                    "start": 1231.28,
                    "end": 1231.36,
                    "word": " So",
                    "probability": 0.7177734375
                },
                {
                    "start": 1231.36,
                    "end": 1232.04,
                    "word": " transformers",
                    "probability": 0.66650390625
                },
                {
                    "start": 1232.04,
                    "end": 1233.98,
                    "word": " kind",
                    "probability": 0.45166015625
                },
                {
                    "start": 1233.98,
                    "end": 1234.08,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 1234.08,
                    "end": 1234.26,
                    "word": " kick",
                    "probability": 0.88916015625
                },
                {
                    "start": 1234.26,
                    "end": 1234.56,
                    "word": " apart",
                    "probability": 0.84912109375
                },
                {
                    "start": 1234.76,
                    "end": 1234.92,
                    "word": " a number",
                    "probability": 0.87841796875
                },
                {
                    "start": 1234.92,
                    "end": 1235.1,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 1235.1,
                    "end": 1235.36,
                    "word": " log",
                    "probability": 0.85595703125
                },
                {
                    "start": 1235.36,
                    "end": 1235.76,
                    "word": " jams",
                    "probability": 0.6707763671875
                },
                {
                    "start": 1235.76,
                    "end": 1236.42,
                    "word": " in",
                    "probability": 0.86083984375
                },
                {
                    "start": 1236.42,
                    "end": 1238.06,
                    "word": " natural",
                    "probability": 0.69677734375
                },
                {
                    "start": 1238.06,
                    "end": 1238.38,
                    "word": " language",
                    "probability": 0.68212890625
                },
                {
                    "start": 1238.38,
                    "end": 1238.74,
                    "word": " modeling",
                    "probability": 0.59375
                },
                {
                    "start": 1238.74,
                    "end": 1238.98,
                    "word": ".",
                    "probability": 0.8955078125
                }
            ]
        },
        {
            "id": 200,
            "seek": 123328,
            "start": 1238.98,
            "end": 1244.32,
            "text": " They give words contextual representations, they model attention, like the relationship between",
            "tokens": [
                51192,
                814,
                976,
                2283,
                35526,
                33358,
                11,
                436,
                2316,
                3202,
                11,
                411,
                264,
                2480,
                1296,
                51472
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22553454496358571,
            "compression_ratio": 1.611353711790393,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1238.98,
                    "end": 1238.98,
                    "word": " They",
                    "probability": 0.720703125
                },
                {
                    "start": 1238.98,
                    "end": 1239.18,
                    "word": " give",
                    "probability": 0.8017578125
                },
                {
                    "start": 1239.18,
                    "end": 1239.94,
                    "word": " words",
                    "probability": 0.60888671875
                },
                {
                    "start": 1239.94,
                    "end": 1240.36,
                    "word": " contextual",
                    "probability": 0.85400390625
                },
                {
                    "start": 1240.36,
                    "end": 1240.9,
                    "word": " representations",
                    "probability": 0.78759765625
                },
                {
                    "start": 1240.9,
                    "end": 1241.82,
                    "word": ",",
                    "probability": 0.4990234375
                },
                {
                    "start": 1241.82,
                    "end": 1242.12,
                    "word": " they",
                    "probability": 0.7265625
                },
                {
                    "start": 1242.12,
                    "end": 1242.66,
                    "word": " model",
                    "probability": 0.84521484375
                },
                {
                    "start": 1242.66,
                    "end": 1243.1,
                    "word": " attention",
                    "probability": 0.80712890625
                },
                {
                    "start": 1243.1,
                    "end": 1243.28,
                    "word": ",",
                    "probability": 0.48291015625
                },
                {
                    "start": 1243.28,
                    "end": 1243.4,
                    "word": " like",
                    "probability": 0.84814453125
                },
                {
                    "start": 1243.4,
                    "end": 1243.5,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 1243.5,
                    "end": 1243.92,
                    "word": " relationship",
                    "probability": 0.8212890625
                },
                {
                    "start": 1243.92,
                    "end": 1244.32,
                    "word": " between",
                    "probability": 0.7841796875
                }
            ]
        },
        {
            "id": 201,
            "seek": 123328,
            "start": 1244.32,
            "end": 1250.26,
            "text": " words in a document, they use these smart partial word representations that allow for",
            "tokens": [
                51472,
                2283,
                294,
                257,
                4166,
                11,
                436,
                764,
                613,
                4069,
                14641,
                1349,
                33358,
                300,
                2089,
                337,
                51772
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22553454496358571,
            "compression_ratio": 1.611353711790393,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1244.32,
                    "end": 1244.84,
                    "word": " words",
                    "probability": 0.7197265625
                },
                {
                    "start": 1244.84,
                    "end": 1245.3,
                    "word": " in",
                    "probability": 0.72998046875
                },
                {
                    "start": 1245.42,
                    "end": 1245.76,
                    "word": " a document",
                    "probability": 0.84716796875
                },
                {
                    "start": 1245.76,
                    "end": 1246.36,
                    "word": ",",
                    "probability": 0.57080078125
                },
                {
                    "start": 1246.36,
                    "end": 1246.62,
                    "word": " they",
                    "probability": 0.72021484375
                },
                {
                    "start": 1246.62,
                    "end": 1247.06,
                    "word": " use",
                    "probability": 0.8212890625
                },
                {
                    "start": 1247.06,
                    "end": 1247.42,
                    "word": " these",
                    "probability": 0.63818359375
                },
                {
                    "start": 1247.42,
                    "end": 1247.8,
                    "word": " smart",
                    "probability": 0.8828125
                },
                {
                    "start": 1247.8,
                    "end": 1248.2,
                    "word": " partial",
                    "probability": 0.82373046875
                },
                {
                    "start": 1248.2,
                    "end": 1248.48,
                    "word": " word",
                    "probability": 0.779296875
                },
                {
                    "start": 1248.48,
                    "end": 1249.04,
                    "word": " representations",
                    "probability": 0.79345703125
                },
                {
                    "start": 1249.04,
                    "end": 1249.64,
                    "word": " that",
                    "probability": 0.8681640625
                },
                {
                    "start": 1249.64,
                    "end": 1249.88,
                    "word": " allow",
                    "probability": 0.865234375
                },
                {
                    "start": 1249.88,
                    "end": 1250.26,
                    "word": " for",
                    "probability": 0.89892578125
                }
            ]
        },
        {
            "id": 202,
            "seek": 126300,
            "start": 1250.26,
            "end": 1254.34,
            "text": " misspellings and just the tower of babble of vernaculars that appear",
            "tokens": [
                50368,
                1713,
                49241,
                1109,
                293,
                445,
                264,
                10567,
                295,
                7564,
                638,
                295,
                35793,
                326,
                42891,
                300,
                4204,
                50564
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23050741984112427,
            "compression_ratio": 1.6539923954372624,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1250.26,
                    "end": 1251.12,
                    "word": " misspellings",
                    "probability": 0.7146809895833334
                },
                {
                    "start": 1251.12,
                    "end": 1251.4,
                    "word": " and",
                    "probability": 0.7578125
                },
                {
                    "start": 1251.4,
                    "end": 1251.92,
                    "word": " just",
                    "probability": 0.57861328125
                },
                {
                    "start": 1251.92,
                    "end": 1252.1,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 1252.1,
                    "end": 1252.32,
                    "word": " tower",
                    "probability": 0.63623046875
                },
                {
                    "start": 1252.32,
                    "end": 1252.5,
                    "word": " of",
                    "probability": 0.8720703125
                },
                {
                    "start": 1252.5,
                    "end": 1252.82,
                    "word": " babble",
                    "probability": 0.7099609375
                },
                {
                    "start": 1252.82,
                    "end": 1253.0,
                    "word": " of",
                    "probability": 0.86767578125
                },
                {
                    "start": 1253.0,
                    "end": 1253.54,
                    "word": " vernaculars",
                    "probability": 0.7763671875
                },
                {
                    "start": 1253.54,
                    "end": 1253.8,
                    "word": " that",
                    "probability": 0.8408203125
                },
                {
                    "start": 1253.8,
                    "end": 1254.34,
                    "word": " appear",
                    "probability": 0.39208984375
                }
            ]
        },
        {
            "id": 203,
            "seek": 126300,
            "start": 1255.8,
            "end": 1260.5,
            "text": " like under the banner of, say, the English language on the internet, and they take advantage",
            "tokens": [
                50652,
                411,
                833,
                264,
                24348,
                295,
                11,
                584,
                11,
                264,
                3669,
                2856,
                322,
                264,
                4705,
                11,
                293,
                436,
                747,
                5002,
                50876
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23050741984112427,
            "compression_ratio": 1.6539923954372624,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1255.8,
                    "end": 1256.36,
                    "word": " like",
                    "probability": 0.31787109375
                },
                {
                    "start": 1256.36,
                    "end": 1256.62,
                    "word": " under",
                    "probability": 0.76953125
                },
                {
                    "start": 1256.62,
                    "end": 1256.74,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 1256.74,
                    "end": 1256.96,
                    "word": " banner",
                    "probability": 0.76513671875
                },
                {
                    "start": 1256.96,
                    "end": 1257.22,
                    "word": " of",
                    "probability": 0.89013671875
                },
                {
                    "start": 1257.22,
                    "end": 1257.38,
                    "word": ",",
                    "probability": 0.321533203125
                },
                {
                    "start": 1257.38,
                    "end": 1257.38,
                    "word": " say",
                    "probability": 0.89453125
                },
                {
                    "start": 1257.38,
                    "end": 1257.42,
                    "word": ",",
                    "probability": 0.89794921875
                },
                {
                    "start": 1257.42,
                    "end": 1257.5,
                    "word": " the",
                    "probability": 0.82080078125
                },
                {
                    "start": 1257.5,
                    "end": 1257.72,
                    "word": " English",
                    "probability": 0.89599609375
                },
                {
                    "start": 1257.72,
                    "end": 1258.1,
                    "word": " language",
                    "probability": 0.6796875
                },
                {
                    "start": 1258.1,
                    "end": 1258.48,
                    "word": " on",
                    "probability": 0.90185546875
                },
                {
                    "start": 1258.48,
                    "end": 1258.86,
                    "word": " the",
                    "probability": 0.7958984375
                },
                {
                    "start": 1258.86,
                    "end": 1259.2,
                    "word": " internet",
                    "probability": 0.467041015625
                },
                {
                    "start": 1259.2,
                    "end": 1259.42,
                    "word": ",",
                    "probability": 0.2135009765625
                },
                {
                    "start": 1259.42,
                    "end": 1259.9,
                    "word": " and",
                    "probability": 0.88818359375
                },
                {
                    "start": 1259.9,
                    "end": 1260.0,
                    "word": " they",
                    "probability": 0.72119140625
                },
                {
                    "start": 1260.0,
                    "end": 1260.16,
                    "word": " take",
                    "probability": 0.81591796875
                },
                {
                    "start": 1260.16,
                    "end": 1260.5,
                    "word": " advantage",
                    "probability": 0.767578125
                }
            ]
        },
        {
            "id": 204,
            "seek": 126300,
            "start": 1260.5,
            "end": 1267.16,
            "text": " of ideas like residual connections and modern optimizers and many of the really good ideas",
            "tokens": [
                50876,
                295,
                3487,
                411,
                27980,
                9271,
                293,
                4363,
                5028,
                22525,
                293,
                867,
                295,
                264,
                534,
                665,
                3487,
                51204
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23050741984112427,
            "compression_ratio": 1.6539923954372624,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1260.5,
                    "end": 1260.82,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1260.82,
                    "end": 1261.26,
                    "word": " ideas",
                    "probability": 0.75537109375
                },
                {
                    "start": 1261.26,
                    "end": 1261.58,
                    "word": " like",
                    "probability": 0.86767578125
                },
                {
                    "start": 1261.58,
                    "end": 1262.22,
                    "word": " residual",
                    "probability": 0.6943359375
                },
                {
                    "start": 1262.22,
                    "end": 1262.76,
                    "word": " connections",
                    "probability": 0.84033203125
                },
                {
                    "start": 1262.76,
                    "end": 1263.3,
                    "word": " and",
                    "probability": 0.79345703125
                },
                {
                    "start": 1263.3,
                    "end": 1264.28,
                    "word": " modern",
                    "probability": 0.8134765625
                },
                {
                    "start": 1264.28,
                    "end": 1265.1,
                    "word": " optimizers",
                    "probability": 0.82861328125
                },
                {
                    "start": 1265.1,
                    "end": 1265.44,
                    "word": " and",
                    "probability": 0.67822265625
                },
                {
                    "start": 1265.44,
                    "end": 1265.8,
                    "word": " many",
                    "probability": 0.8115234375
                },
                {
                    "start": 1265.8,
                    "end": 1265.94,
                    "word": " of",
                    "probability": 0.896484375
                },
                {
                    "start": 1265.94,
                    "end": 1266.26,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 1266.26,
                    "end": 1266.62,
                    "word": " really",
                    "probability": 0.75439453125
                },
                {
                    "start": 1266.62,
                    "end": 1266.84,
                    "word": " good",
                    "probability": 0.8798828125
                },
                {
                    "start": 1266.84,
                    "end": 1267.16,
                    "word": " ideas",
                    "probability": 0.765625
                }
            ]
        },
        {
            "id": 205,
            "seek": 126300,
            "start": 1267.16,
            "end": 1271.98,
            "text": " that have come out of the neural network revolution. So these are all reasons why",
            "tokens": [
                51204,
                300,
                362,
                808,
                484,
                295,
                264,
                18161,
                3209,
                8894,
                13,
                407,
                613,
                366,
                439,
                4112,
                983,
                51444
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23050741984112427,
            "compression_ratio": 1.6539923954372624,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1267.16,
                    "end": 1267.32,
                    "word": " that",
                    "probability": 0.88525390625
                },
                {
                    "start": 1267.32,
                    "end": 1267.42,
                    "word": " have",
                    "probability": 0.8564453125
                },
                {
                    "start": 1267.42,
                    "end": 1267.58,
                    "word": " come",
                    "probability": 0.7998046875
                },
                {
                    "start": 1267.58,
                    "end": 1267.82,
                    "word": " out",
                    "probability": 0.8173828125
                },
                {
                    "start": 1267.82,
                    "end": 1268.22,
                    "word": " of",
                    "probability": 0.8955078125
                },
                {
                    "start": 1268.22,
                    "end": 1268.46,
                    "word": " the",
                    "probability": 0.79833984375
                },
                {
                    "start": 1268.46,
                    "end": 1268.64,
                    "word": " neural",
                    "probability": 0.8037109375
                },
                {
                    "start": 1268.64,
                    "end": 1268.92,
                    "word": " network",
                    "probability": 0.8408203125
                },
                {
                    "start": 1268.92,
                    "end": 1269.58,
                    "word": " revolution",
                    "probability": 0.76318359375
                },
                {
                    "start": 1269.58,
                    "end": 1270.14,
                    "word": ".",
                    "probability": 0.9404296875
                },
                {
                    "start": 1270.6,
                    "end": 1270.76,
                    "word": " So",
                    "probability": 0.8232421875
                },
                {
                    "start": 1270.76,
                    "end": 1270.88,
                    "word": " these",
                    "probability": 0.62939453125
                },
                {
                    "start": 1270.88,
                    "end": 1270.98,
                    "word": " are",
                    "probability": 0.8935546875
                },
                {
                    "start": 1270.98,
                    "end": 1271.2,
                    "word": " all",
                    "probability": 0.8984375
                },
                {
                    "start": 1271.2,
                    "end": 1271.82,
                    "word": " reasons",
                    "probability": 0.80517578125
                },
                {
                    "start": 1271.82,
                    "end": 1271.98,
                    "word": " why",
                    "probability": 0.810546875
                }
            ]
        },
        {
            "id": 206,
            "seek": 126300,
            "start": 1271.98,
            "end": 1281.54,
            "text": " we wanted to test their applicability to phishing detection. Okay, so if you want to get into detail",
            "tokens": [
                51444,
                321,
                1415,
                281,
                1500,
                641,
                2580,
                2310,
                281,
                903,
                3807,
                17784,
                13,
                1033,
                11,
                370,
                498,
                291,
                528,
                281,
                483,
                666,
                2607,
                51852
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23050741984112427,
            "compression_ratio": 1.6539923954372624,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1271.98,
                    "end": 1272.08,
                    "word": " we",
                    "probability": 0.90869140625
                },
                {
                    "start": 1272.08,
                    "end": 1272.24,
                    "word": " wanted",
                    "probability": 0.8154296875
                },
                {
                    "start": 1272.24,
                    "end": 1272.4,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 1272.4,
                    "end": 1272.66,
                    "word": " test",
                    "probability": 0.81591796875
                },
                {
                    "start": 1272.66,
                    "end": 1273.04,
                    "word": " their",
                    "probability": 0.90283203125
                },
                {
                    "start": 1273.04,
                    "end": 1273.9,
                    "word": " applicability",
                    "probability": 0.925048828125
                },
                {
                    "start": 1273.9,
                    "end": 1274.22,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 1274.22,
                    "end": 1274.68,
                    "word": " phishing",
                    "probability": 0.848388671875
                },
                {
                    "start": 1274.68,
                    "end": 1275.08,
                    "word": " detection",
                    "probability": 0.779296875
                },
                {
                    "start": 1275.08,
                    "end": 1275.64,
                    "word": ".",
                    "probability": 0.9912109375
                },
                {
                    "start": 1278.44,
                    "end": 1278.72,
                    "word": " Okay",
                    "probability": 0.59423828125
                },
                {
                    "start": 1278.72,
                    "end": 1279.0,
                    "word": ",",
                    "probability": 0.7177734375
                },
                {
                    "start": 1279.0,
                    "end": 1279.28,
                    "word": " so",
                    "probability": 0.896484375
                },
                {
                    "start": 1279.28,
                    "end": 1280.48,
                    "word": " if",
                    "probability": 0.83447265625
                },
                {
                    "start": 1280.48,
                    "end": 1280.56,
                    "word": " you",
                    "probability": 0.8935546875
                },
                {
                    "start": 1280.56,
                    "end": 1280.66,
                    "word": " want",
                    "probability": 0.7685546875
                },
                {
                    "start": 1280.66,
                    "end": 1280.74,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1280.74,
                    "end": 1280.88,
                    "word": " get",
                    "probability": 0.8935546875
                },
                {
                    "start": 1280.88,
                    "end": 1281.08,
                    "word": " into",
                    "probability": 0.73583984375
                },
                {
                    "start": 1281.08,
                    "end": 1281.54,
                    "word": " detail",
                    "probability": 0.84033203125
                }
            ]
        },
        {
            "id": 207,
            "seek": 128710,
            "start": 1281.54,
            "end": 1287.78,
            "text": " about how transformers work, I'd recommend this blog post by Jay Alomar. It's where I cribbed this",
            "tokens": [
                50368,
                466,
                577,
                4088,
                433,
                589,
                11,
                286,
                1116,
                2748,
                341,
                6968,
                2183,
                538,
                11146,
                967,
                298,
                289,
                13,
                467,
                311,
                689,
                286,
                47163,
                2883,
                341,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22194602086462759,
            "compression_ratio": 1.6491228070175439,
            "no_speech_prob": 0.0004875659942626953,
            "words": [
                {
                    "start": 1281.54,
                    "end": 1281.84,
                    "word": " about",
                    "probability": 0.477783203125
                },
                {
                    "start": 1281.84,
                    "end": 1282.1,
                    "word": " how",
                    "probability": 0.87158203125
                },
                {
                    "start": 1282.1,
                    "end": 1282.58,
                    "word": " transformers",
                    "probability": 0.89453125
                },
                {
                    "start": 1282.58,
                    "end": 1282.78,
                    "word": " work",
                    "probability": 0.7958984375
                },
                {
                    "start": 1282.78,
                    "end": 1282.86,
                    "word": ",",
                    "probability": 0.53857421875
                },
                {
                    "start": 1282.86,
                    "end": 1283.02,
                    "word": " I'd",
                    "probability": 0.96435546875
                },
                {
                    "start": 1283.02,
                    "end": 1283.28,
                    "word": " recommend",
                    "probability": 0.69482421875
                },
                {
                    "start": 1283.28,
                    "end": 1283.5,
                    "word": " this",
                    "probability": 0.87939453125
                },
                {
                    "start": 1283.5,
                    "end": 1283.74,
                    "word": " blog",
                    "probability": 0.83642578125
                },
                {
                    "start": 1283.74,
                    "end": 1284.14,
                    "word": " post",
                    "probability": 0.80859375
                },
                {
                    "start": 1284.14,
                    "end": 1284.44,
                    "word": " by",
                    "probability": 0.89697265625
                },
                {
                    "start": 1284.44,
                    "end": 1284.88,
                    "word": " Jay",
                    "probability": 0.6708984375
                },
                {
                    "start": 1284.88,
                    "end": 1286.04,
                    "word": " Alomar",
                    "probability": 0.741455078125
                },
                {
                    "start": 1286.04,
                    "end": 1286.38,
                    "word": ".",
                    "probability": 0.5458984375
                },
                {
                    "start": 1286.38,
                    "end": 1286.6,
                    "word": " It's",
                    "probability": 0.805908203125
                },
                {
                    "start": 1286.6,
                    "end": 1286.76,
                    "word": " where",
                    "probability": 0.90771484375
                },
                {
                    "start": 1286.76,
                    "end": 1286.96,
                    "word": " I",
                    "probability": 0.9853515625
                },
                {
                    "start": 1286.96,
                    "end": 1287.4,
                    "word": " cribbed",
                    "probability": 0.847900390625
                },
                {
                    "start": 1287.4,
                    "end": 1287.78,
                    "word": " this",
                    "probability": 0.875
                }
            ]
        },
        {
            "id": 208,
            "seek": 128710,
            "start": 1288.34,
            "end": 1294.36,
            "text": " screenshot here. I'm not going to get into the details of the series of matrix multiplications",
            "tokens": [
                50704,
                27712,
                510,
                13,
                286,
                478,
                406,
                516,
                281,
                483,
                666,
                264,
                4365,
                295,
                264,
                2638,
                295,
                8141,
                17596,
                763,
                51004
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22194602086462759,
            "compression_ratio": 1.6491228070175439,
            "no_speech_prob": 0.0004875659942626953,
            "words": [
                {
                    "start": 1288.34,
                    "end": 1288.82,
                    "word": " screenshot",
                    "probability": 0.8076171875
                },
                {
                    "start": 1288.82,
                    "end": 1289.3,
                    "word": " here",
                    "probability": 0.76953125
                },
                {
                    "start": 1289.3,
                    "end": 1289.48,
                    "word": ".",
                    "probability": 0.8759765625
                },
                {
                    "start": 1289.48,
                    "end": 1289.72,
                    "word": " I'm",
                    "probability": 0.9755859375
                },
                {
                    "start": 1289.72,
                    "end": 1289.88,
                    "word": " not",
                    "probability": 0.8955078125
                },
                {
                    "start": 1289.88,
                    "end": 1289.98,
                    "word": " going",
                    "probability": 0.66796875
                },
                {
                    "start": 1289.98,
                    "end": 1290.02,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 1290.02,
                    "end": 1290.18,
                    "word": " get",
                    "probability": 0.90478515625
                },
                {
                    "start": 1290.18,
                    "end": 1290.58,
                    "word": " into",
                    "probability": 0.72705078125
                },
                {
                    "start": 1290.58,
                    "end": 1291.34,
                    "word": " the",
                    "probability": 0.81884765625
                },
                {
                    "start": 1291.34,
                    "end": 1291.74,
                    "word": " details",
                    "probability": 0.6826171875
                },
                {
                    "start": 1291.74,
                    "end": 1292.32,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1292.32,
                    "end": 1292.76,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 1292.76,
                    "end": 1293.22,
                    "word": " series",
                    "probability": 0.82080078125
                },
                {
                    "start": 1293.22,
                    "end": 1293.38,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 1293.38,
                    "end": 1293.64,
                    "word": " matrix",
                    "probability": 0.83642578125
                },
                {
                    "start": 1293.64,
                    "end": 1294.36,
                    "word": " multiplications",
                    "probability": 0.888916015625
                }
            ]
        },
        {
            "id": 209,
            "seek": 128710,
            "start": 1294.36,
            "end": 1301.14,
            "text": " that comprise summations and various linear algebra and operations that comprise the",
            "tokens": [
                51004,
                300,
                16802,
                908,
                8367,
                763,
                293,
                3683,
                8213,
                21989,
                293,
                7705,
                300,
                16802,
                908,
                264,
                51336
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22194602086462759,
            "compression_ratio": 1.6491228070175439,
            "no_speech_prob": 0.0004875659942626953,
            "words": [
                {
                    "start": 1294.36,
                    "end": 1294.62,
                    "word": " that",
                    "probability": 0.85546875
                },
                {
                    "start": 1294.62,
                    "end": 1295.26,
                    "word": " comprise",
                    "probability": 0.840087890625
                },
                {
                    "start": 1295.26,
                    "end": 1297.5,
                    "word": " summations",
                    "probability": 0.5621337890625
                },
                {
                    "start": 1297.5,
                    "end": 1297.86,
                    "word": " and",
                    "probability": 0.791015625
                },
                {
                    "start": 1297.86,
                    "end": 1298.44,
                    "word": " various",
                    "probability": 0.83154296875
                },
                {
                    "start": 1298.44,
                    "end": 1299.32,
                    "word": " linear",
                    "probability": 0.86279296875
                },
                {
                    "start": 1299.32,
                    "end": 1299.66,
                    "word": " algebra",
                    "probability": 0.7294921875
                },
                {
                    "start": 1299.66,
                    "end": 1299.88,
                    "word": " and",
                    "probability": 0.8134765625
                },
                {
                    "start": 1299.88,
                    "end": 1300.28,
                    "word": " operations",
                    "probability": 0.88818359375
                },
                {
                    "start": 1300.28,
                    "end": 1300.54,
                    "word": " that",
                    "probability": 0.86865234375
                },
                {
                    "start": 1300.54,
                    "end": 1301.04,
                    "word": " comprise",
                    "probability": 0.84912109375
                },
                {
                    "start": 1301.04,
                    "end": 1301.14,
                    "word": " the",
                    "probability": 0.269287109375
                }
            ]
        },
        {
            "id": 210,
            "seek": 128710,
            "start": 1301.14,
            "end": 1305.64,
            "text": " transformer block. I just want to give a little bit of intuition here before passing the mic over",
            "tokens": [
                51336,
                31782,
                3461,
                13,
                286,
                445,
                528,
                281,
                976,
                257,
                707,
                857,
                295,
                24002,
                510,
                949,
                8437,
                264,
                3123,
                670,
                51568
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22194602086462759,
            "compression_ratio": 1.6491228070175439,
            "no_speech_prob": 0.0004875659942626953,
            "words": [
                {
                    "start": 1301.14,
                    "end": 1301.42,
                    "word": " transformer",
                    "probability": 0.70556640625
                },
                {
                    "start": 1301.42,
                    "end": 1301.6,
                    "word": " block",
                    "probability": 0.80810546875
                },
                {
                    "start": 1301.6,
                    "end": 1301.66,
                    "word": ".",
                    "probability": 0.8095703125
                },
                {
                    "start": 1301.66,
                    "end": 1301.7,
                    "word": " I",
                    "probability": 0.99462890625
                },
                {
                    "start": 1301.7,
                    "end": 1301.82,
                    "word": " just",
                    "probability": 0.8818359375
                },
                {
                    "start": 1301.82,
                    "end": 1301.92,
                    "word": " want",
                    "probability": 0.80859375
                },
                {
                    "start": 1301.92,
                    "end": 1301.98,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1301.98,
                    "end": 1302.1,
                    "word": " give",
                    "probability": 0.8046875
                },
                {
                    "start": 1302.32,
                    "end": 1302.46,
                    "word": " a little",
                    "probability": 0.8076171875
                },
                {
                    "start": 1302.46,
                    "end": 1302.58,
                    "word": " bit",
                    "probability": 0.8984375
                },
                {
                    "start": 1302.58,
                    "end": 1302.68,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 1302.68,
                    "end": 1303.0,
                    "word": " intuition",
                    "probability": 0.80908203125
                },
                {
                    "start": 1303.0,
                    "end": 1303.34,
                    "word": " here",
                    "probability": 0.802734375
                },
                {
                    "start": 1303.34,
                    "end": 1303.66,
                    "word": " before",
                    "probability": 0.7138671875
                },
                {
                    "start": 1303.66,
                    "end": 1304.04,
                    "word": " passing",
                    "probability": 0.76220703125
                },
                {
                    "start": 1304.04,
                    "end": 1305.34,
                    "word": " the",
                    "probability": 0.76171875
                },
                {
                    "start": 1305.34,
                    "end": 1305.48,
                    "word": " mic",
                    "probability": 0.82421875
                },
                {
                    "start": 1305.48,
                    "end": 1305.64,
                    "word": " over",
                    "probability": 0.8828125
                }
            ]
        },
        {
            "id": 211,
            "seek": 131286,
            "start": 1305.64,
            "end": 1312.3,
            "text": " to Younghoo. The basic idea behind a transformer block, which is kind of a Lego block,",
            "tokens": [
                50376,
                281,
                8160,
                19069,
                13,
                440,
                3875,
                1558,
                2261,
                257,
                31782,
                3461,
                11,
                597,
                307,
                733,
                295,
                257,
                28761,
                3461,
                11,
                50676
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2415226119312834,
            "compression_ratio": 1.6419213973799127,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 1305.64,
                    "end": 1305.78,
                    "word": " to",
                    "probability": 0.421142578125
                },
                {
                    "start": 1305.78,
                    "end": 1306.46,
                    "word": " Younghoo",
                    "probability": 0.3636474609375
                },
                {
                    "start": 1306.46,
                    "end": 1306.98,
                    "word": ".",
                    "probability": 0.371826171875
                },
                {
                    "start": 1307.24,
                    "end": 1307.76,
                    "word": " The",
                    "probability": 0.779296875
                },
                {
                    "start": 1307.76,
                    "end": 1308.1,
                    "word": " basic",
                    "probability": 0.90869140625
                },
                {
                    "start": 1308.1,
                    "end": 1308.38,
                    "word": " idea",
                    "probability": 0.7880859375
                },
                {
                    "start": 1308.38,
                    "end": 1308.78,
                    "word": " behind",
                    "probability": 0.86083984375
                },
                {
                    "start": 1309.94,
                    "end": 1310.34,
                    "word": " a transformer",
                    "probability": 0.7119140625
                },
                {
                    "start": 1310.34,
                    "end": 1310.76,
                    "word": " block",
                    "probability": 0.82080078125
                },
                {
                    "start": 1310.76,
                    "end": 1310.86,
                    "word": ",",
                    "probability": 0.71240234375
                },
                {
                    "start": 1310.86,
                    "end": 1310.98,
                    "word": " which",
                    "probability": 0.92041015625
                },
                {
                    "start": 1310.98,
                    "end": 1311.08,
                    "word": " is",
                    "probability": 0.904296875
                },
                {
                    "start": 1311.08,
                    "end": 1311.2,
                    "word": " kind",
                    "probability": 0.7314453125
                },
                {
                    "start": 1311.2,
                    "end": 1311.24,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 1311.38,
                    "end": 1311.56,
                    "word": " a Lego",
                    "probability": 0.58447265625
                },
                {
                    "start": 1311.56,
                    "end": 1312.02,
                    "word": " block",
                    "probability": 0.80615234375
                },
                {
                    "start": 1312.02,
                    "end": 1312.3,
                    "word": ",",
                    "probability": 0.3974609375
                }
            ]
        },
        {
            "id": 212,
            "seek": 131286,
            "start": 1312.86,
            "end": 1319.54,
            "text": " out of which you build a transformer-based neural network, is we're taking in a sequence of words.",
            "tokens": [
                50736,
                484,
                295,
                597,
                291,
                1322,
                257,
                31782,
                12,
                6032,
                18161,
                3209,
                11,
                307,
                321,
                434,
                1940,
                294,
                257,
                8310,
                295,
                2283,
                13,
                51032
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2415226119312834,
            "compression_ratio": 1.6419213973799127,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 1312.86,
                    "end": 1313.38,
                    "word": " out",
                    "probability": 0.7861328125
                },
                {
                    "start": 1313.38,
                    "end": 1313.5,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 1313.5,
                    "end": 1313.68,
                    "word": " which",
                    "probability": 0.91064453125
                },
                {
                    "start": 1313.68,
                    "end": 1313.82,
                    "word": " you",
                    "probability": 0.890625
                },
                {
                    "start": 1313.82,
                    "end": 1313.98,
                    "word": " build",
                    "probability": 0.87890625
                },
                {
                    "start": 1314.12,
                    "end": 1314.58,
                    "word": " a transformer",
                    "probability": 0.78759765625
                },
                {
                    "start": 1314.58,
                    "end": 1315.2,
                    "word": "-based",
                    "probability": 0.626708984375
                },
                {
                    "start": 1315.2,
                    "end": 1315.7,
                    "word": " neural",
                    "probability": 0.91552734375
                },
                {
                    "start": 1315.7,
                    "end": 1316.14,
                    "word": " network",
                    "probability": 0.828125
                },
                {
                    "start": 1316.14,
                    "end": 1316.32,
                    "word": ",",
                    "probability": 0.71240234375
                },
                {
                    "start": 1316.32,
                    "end": 1316.66,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 1316.66,
                    "end": 1317.46,
                    "word": " we're",
                    "probability": 0.863037109375
                },
                {
                    "start": 1317.46,
                    "end": 1317.72,
                    "word": " taking",
                    "probability": 0.80078125
                },
                {
                    "start": 1317.72,
                    "end": 1318.04,
                    "word": " in",
                    "probability": 0.89208984375
                },
                {
                    "start": 1318.3,
                    "end": 1318.52,
                    "word": " a sequence",
                    "probability": 0.89501953125
                },
                {
                    "start": 1318.52,
                    "end": 1318.74,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1318.74,
                    "end": 1319.02,
                    "word": " words",
                    "probability": 0.728515625
                },
                {
                    "start": 1319.02,
                    "end": 1319.54,
                    "word": ".",
                    "probability": 0.68896484375
                }
            ]
        },
        {
            "id": 213,
            "seek": 131286,
            "start": 1320.14,
            "end": 1324.06,
            "text": " This diagram shows a very simple example where we're just taking a sequence of two words. Typically",
            "tokens": [
                51100,
                639,
                10686,
                3110,
                257,
                588,
                2199,
                1365,
                689,
                321,
                434,
                445,
                1940,
                257,
                8310,
                295,
                732,
                2283,
                13,
                23129,
                51284
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2415226119312834,
            "compression_ratio": 1.6419213973799127,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 1320.14,
                    "end": 1320.66,
                    "word": " This",
                    "probability": 0.80712890625
                },
                {
                    "start": 1320.66,
                    "end": 1321.0,
                    "word": " diagram",
                    "probability": 0.74560546875
                },
                {
                    "start": 1321.0,
                    "end": 1321.3,
                    "word": " shows",
                    "probability": 0.8125
                },
                {
                    "start": 1321.42,
                    "end": 1321.54,
                    "word": " a very",
                    "probability": 0.734375
                },
                {
                    "start": 1321.54,
                    "end": 1321.82,
                    "word": " simple",
                    "probability": 0.912109375
                },
                {
                    "start": 1321.82,
                    "end": 1322.16,
                    "word": " example",
                    "probability": 0.92724609375
                },
                {
                    "start": 1322.16,
                    "end": 1322.32,
                    "word": " where",
                    "probability": 0.76123046875
                },
                {
                    "start": 1322.32,
                    "end": 1322.44,
                    "word": " we're",
                    "probability": 0.729736328125
                },
                {
                    "start": 1322.44,
                    "end": 1322.56,
                    "word": " just",
                    "probability": 0.88134765625
                },
                {
                    "start": 1322.56,
                    "end": 1322.72,
                    "word": " taking",
                    "probability": 0.79443359375
                },
                {
                    "start": 1322.84,
                    "end": 1323.06,
                    "word": " a sequence",
                    "probability": 0.8974609375
                },
                {
                    "start": 1323.06,
                    "end": 1323.2,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1323.2,
                    "end": 1323.38,
                    "word": " two",
                    "probability": 0.8115234375
                },
                {
                    "start": 1323.38,
                    "end": 1323.66,
                    "word": " words",
                    "probability": 0.732421875
                },
                {
                    "start": 1323.66,
                    "end": 1323.84,
                    "word": ".",
                    "probability": 0.66845703125
                },
                {
                    "start": 1323.84,
                    "end": 1324.06,
                    "word": " Typically",
                    "probability": 0.841796875
                }
            ]
        },
        {
            "id": 214,
            "seek": 131286,
            "start": 1324.06,
            "end": 1331.4,
            "text": " we've taken a larger window, like 512 words. We pass them into the block, and the way they",
            "tokens": [
                51284,
                321,
                600,
                2726,
                257,
                4833,
                4910,
                11,
                411,
                1025,
                4762,
                2283,
                13,
                492,
                1320,
                552,
                666,
                264,
                3461,
                11,
                293,
                264,
                636,
                436,
                51648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2415226119312834,
            "compression_ratio": 1.6419213973799127,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 1324.06,
                    "end": 1324.26,
                    "word": " we've",
                    "probability": 0.515869140625
                },
                {
                    "start": 1324.26,
                    "end": 1324.48,
                    "word": " taken",
                    "probability": 0.7431640625
                },
                {
                    "start": 1325.12,
                    "end": 1325.36,
                    "word": " a larger",
                    "probability": 0.81689453125
                },
                {
                    "start": 1325.36,
                    "end": 1325.64,
                    "word": " window",
                    "probability": 0.841796875
                },
                {
                    "start": 1325.64,
                    "end": 1325.74,
                    "word": ",",
                    "probability": 0.66162109375
                },
                {
                    "start": 1325.74,
                    "end": 1325.88,
                    "word": " like",
                    "probability": 0.87353515625
                },
                {
                    "start": 1325.88,
                    "end": 1326.54,
                    "word": " 512",
                    "probability": 0.94921875
                },
                {
                    "start": 1326.54,
                    "end": 1326.92,
                    "word": " words",
                    "probability": 0.73974609375
                },
                {
                    "start": 1326.92,
                    "end": 1327.44,
                    "word": ".",
                    "probability": 0.736328125
                },
                {
                    "start": 1327.58,
                    "end": 1328.0,
                    "word": " We",
                    "probability": 0.8955078125
                },
                {
                    "start": 1328.0,
                    "end": 1328.48,
                    "word": " pass",
                    "probability": 0.79150390625
                },
                {
                    "start": 1328.48,
                    "end": 1328.6,
                    "word": " them",
                    "probability": 0.72998046875
                },
                {
                    "start": 1328.6,
                    "end": 1328.74,
                    "word": " into",
                    "probability": 0.7314453125
                },
                {
                    "start": 1328.74,
                    "end": 1328.88,
                    "word": " the",
                    "probability": 0.81884765625
                },
                {
                    "start": 1328.88,
                    "end": 1329.16,
                    "word": " block",
                    "probability": 0.8203125
                },
                {
                    "start": 1329.16,
                    "end": 1329.48,
                    "word": ",",
                    "probability": 0.211181640625
                },
                {
                    "start": 1329.48,
                    "end": 1330.72,
                    "word": " and",
                    "probability": 0.8916015625
                },
                {
                    "start": 1330.72,
                    "end": 1330.96,
                    "word": " the",
                    "probability": 0.79443359375
                },
                {
                    "start": 1330.96,
                    "end": 1331.28,
                    "word": " way",
                    "probability": 0.89501953125
                },
                {
                    "start": 1331.28,
                    "end": 1331.4,
                    "word": " they",
                    "probability": 0.7060546875
                }
            ]
        },
        {
            "id": 215,
            "seek": 133690,
            "start": 1331.4,
            "end": 1340.76,
            "text": " get passed in is not just as entries in a term vector matrix, but actually as vectors themselves.",
            "tokens": [
                50364,
                483,
                4678,
                294,
                307,
                406,
                445,
                382,
                23041,
                294,
                257,
                1433,
                8062,
                8141,
                11,
                457,
                767,
                382,
                18875,
                2969,
                13,
                50828
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22321429281007676,
            "compression_ratio": 1.6171428571428572,
            "no_speech_prob": 0.0012836456298828125,
            "words": [
                {
                    "start": 1331.4,
                    "end": 1331.5,
                    "word": " get",
                    "probability": 0.30615234375
                },
                {
                    "start": 1331.5,
                    "end": 1331.8,
                    "word": " passed",
                    "probability": 0.7119140625
                },
                {
                    "start": 1331.8,
                    "end": 1332.04,
                    "word": " in",
                    "probability": 0.90673828125
                },
                {
                    "start": 1332.04,
                    "end": 1333.06,
                    "word": " is",
                    "probability": 0.7578125
                },
                {
                    "start": 1333.06,
                    "end": 1334.76,
                    "word": " not",
                    "probability": 0.2164306640625
                },
                {
                    "start": 1334.76,
                    "end": 1335.06,
                    "word": " just",
                    "probability": 0.88330078125
                },
                {
                    "start": 1335.06,
                    "end": 1337.74,
                    "word": " as",
                    "probability": 0.80712890625
                },
                {
                    "start": 1337.74,
                    "end": 1338.08,
                    "word": " entries",
                    "probability": 0.81103515625
                },
                {
                    "start": 1338.08,
                    "end": 1338.28,
                    "word": " in",
                    "probability": 0.82861328125
                },
                {
                    "start": 1338.4,
                    "end": 1338.6,
                    "word": " a term",
                    "probability": 0.7138671875
                },
                {
                    "start": 1338.6,
                    "end": 1338.86,
                    "word": " vector",
                    "probability": 0.83251953125
                },
                {
                    "start": 1338.86,
                    "end": 1339.12,
                    "word": " matrix",
                    "probability": 0.830078125
                },
                {
                    "start": 1339.12,
                    "end": 1339.22,
                    "word": ",",
                    "probability": 0.568359375
                },
                {
                    "start": 1339.22,
                    "end": 1339.3,
                    "word": " but",
                    "probability": 0.8896484375
                },
                {
                    "start": 1339.3,
                    "end": 1339.52,
                    "word": " actually",
                    "probability": 0.77197265625
                },
                {
                    "start": 1339.52,
                    "end": 1339.7,
                    "word": " as",
                    "probability": 0.875
                },
                {
                    "start": 1339.7,
                    "end": 1340.02,
                    "word": " vectors",
                    "probability": 0.892578125
                },
                {
                    "start": 1340.02,
                    "end": 1340.56,
                    "word": " themselves",
                    "probability": 0.75390625
                },
                {
                    "start": 1340.56,
                    "end": 1340.76,
                    "word": ".",
                    "probability": 0.78173828125
                }
            ]
        },
        {
            "id": 216,
            "seek": 133690,
            "start": 1340.76,
            "end": 1347.92,
            "text": " The words get a vector representation. These are known as embeddings. We pass them both into the",
            "tokens": [
                50828,
                440,
                2283,
                483,
                257,
                8062,
                10290,
                13,
                1981,
                366,
                2570,
                382,
                12240,
                29432,
                13,
                492,
                1320,
                552,
                1293,
                666,
                264,
                51184
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22321429281007676,
            "compression_ratio": 1.6171428571428572,
            "no_speech_prob": 0.0012836456298828125,
            "words": [
                {
                    "start": 1340.76,
                    "end": 1340.88,
                    "word": " The",
                    "probability": 0.2117919921875
                },
                {
                    "start": 1340.88,
                    "end": 1341.06,
                    "word": " words",
                    "probability": 0.66943359375
                },
                {
                    "start": 1341.06,
                    "end": 1341.22,
                    "word": " get",
                    "probability": 0.724609375
                },
                {
                    "start": 1341.32,
                    "end": 1341.52,
                    "word": " a vector",
                    "probability": 0.921875
                },
                {
                    "start": 1341.52,
                    "end": 1341.96,
                    "word": " representation",
                    "probability": 0.72412109375
                },
                {
                    "start": 1341.96,
                    "end": 1342.14,
                    "word": ".",
                    "probability": 0.39013671875
                },
                {
                    "start": 1342.14,
                    "end": 1342.2,
                    "word": " These",
                    "probability": 0.90625
                },
                {
                    "start": 1342.2,
                    "end": 1342.28,
                    "word": " are",
                    "probability": 0.89599609375
                },
                {
                    "start": 1342.28,
                    "end": 1342.54,
                    "word": " known",
                    "probability": 0.7607421875
                },
                {
                    "start": 1342.54,
                    "end": 1342.66,
                    "word": " as",
                    "probability": 0.87646484375
                },
                {
                    "start": 1342.66,
                    "end": 1343.14,
                    "word": " embeddings",
                    "probability": 0.886474609375
                },
                {
                    "start": 1343.14,
                    "end": 1343.66,
                    "word": ".",
                    "probability": 0.74609375
                },
                {
                    "start": 1345.92,
                    "end": 1346.44,
                    "word": " We",
                    "probability": 0.29150390625
                },
                {
                    "start": 1346.44,
                    "end": 1347.18,
                    "word": " pass",
                    "probability": 0.72265625
                },
                {
                    "start": 1347.18,
                    "end": 1347.32,
                    "word": " them",
                    "probability": 0.73193359375
                },
                {
                    "start": 1347.32,
                    "end": 1347.58,
                    "word": " both",
                    "probability": 0.8203125
                },
                {
                    "start": 1347.58,
                    "end": 1347.78,
                    "word": " into",
                    "probability": 0.73974609375
                },
                {
                    "start": 1347.78,
                    "end": 1347.92,
                    "word": " the",
                    "probability": 0.81787109375
                }
            ]
        },
        {
            "id": 217,
            "seek": 133690,
            "start": 1347.92,
            "end": 1355.44,
            "text": " transformer network. The first thing that the transformer network does to our input word",
            "tokens": [
                51184,
                31782,
                3209,
                13,
                440,
                700,
                551,
                300,
                264,
                31782,
                3209,
                775,
                281,
                527,
                4846,
                1349,
                51564
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22321429281007676,
            "compression_ratio": 1.6171428571428572,
            "no_speech_prob": 0.0012836456298828125,
            "words": [
                {
                    "start": 1347.92,
                    "end": 1348.24,
                    "word": " transformer",
                    "probability": 0.71630859375
                },
                {
                    "start": 1348.24,
                    "end": 1348.58,
                    "word": " network",
                    "probability": 0.8291015625
                },
                {
                    "start": 1348.58,
                    "end": 1349.1,
                    "word": ".",
                    "probability": 0.98779296875
                },
                {
                    "start": 1350.0,
                    "end": 1350.34,
                    "word": " The",
                    "probability": 0.81640625
                },
                {
                    "start": 1350.34,
                    "end": 1351.14,
                    "word": " first",
                    "probability": 0.8134765625
                },
                {
                    "start": 1351.14,
                    "end": 1351.42,
                    "word": " thing",
                    "probability": 0.7880859375
                },
                {
                    "start": 1351.42,
                    "end": 1351.56,
                    "word": " that",
                    "probability": 0.85986328125
                },
                {
                    "start": 1351.56,
                    "end": 1351.68,
                    "word": " the",
                    "probability": 0.81982421875
                },
                {
                    "start": 1351.68,
                    "end": 1352.0,
                    "word": " transformer",
                    "probability": 0.810546875
                },
                {
                    "start": 1352.0,
                    "end": 1352.32,
                    "word": " network",
                    "probability": 0.8212890625
                },
                {
                    "start": 1352.32,
                    "end": 1353.04,
                    "word": " does",
                    "probability": 0.93505859375
                },
                {
                    "start": 1353.04,
                    "end": 1353.84,
                    "word": " to",
                    "probability": 0.900390625
                },
                {
                    "start": 1353.84,
                    "end": 1354.74,
                    "word": " our",
                    "probability": 0.7861328125
                },
                {
                    "start": 1354.74,
                    "end": 1355.14,
                    "word": " input",
                    "probability": 0.88037109375
                },
                {
                    "start": 1355.14,
                    "end": 1355.44,
                    "word": " word",
                    "probability": 0.79150390625
                }
            ]
        },
        {
            "id": 218,
            "seek": 136570,
            "start": 1355.44,
            "end": 1365.52,
            "text": " sequence is model the attention relationships between the words. In a two-word case, it's a",
            "tokens": [
                50364,
                8310,
                307,
                2316,
                264,
                3202,
                6159,
                1296,
                264,
                2283,
                13,
                682,
                257,
                732,
                12,
                7462,
                1389,
                11,
                309,
                311,
                257,
                50860
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26285601492169536,
            "compression_ratio": 1.5603864734299517,
            "no_speech_prob": 0.010009765625,
            "words": [
                {
                    "start": 1355.44,
                    "end": 1355.84,
                    "word": " sequence",
                    "probability": 0.346923828125
                },
                {
                    "start": 1355.84,
                    "end": 1356.3,
                    "word": " is",
                    "probability": 0.83935546875
                },
                {
                    "start": 1356.3,
                    "end": 1357.2,
                    "word": " model",
                    "probability": 0.728515625
                },
                {
                    "start": 1357.2,
                    "end": 1357.66,
                    "word": " the",
                    "probability": 0.77197265625
                },
                {
                    "start": 1357.66,
                    "end": 1358.54,
                    "word": " attention",
                    "probability": 0.368408203125
                },
                {
                    "start": 1358.54,
                    "end": 1360.62,
                    "word": " relationships",
                    "probability": 0.7060546875
                },
                {
                    "start": 1360.62,
                    "end": 1361.1,
                    "word": " between",
                    "probability": 0.787109375
                },
                {
                    "start": 1361.1,
                    "end": 1361.8,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1361.8,
                    "end": 1362.24,
                    "word": " words",
                    "probability": 0.71142578125
                },
                {
                    "start": 1362.24,
                    "end": 1362.92,
                    "word": ".",
                    "probability": 0.9287109375
                },
                {
                    "start": 1363.32,
                    "end": 1363.98,
                    "word": " In",
                    "probability": 0.05670166015625
                },
                {
                    "start": 1364.66,
                    "end": 1364.8,
                    "word": " a two",
                    "probability": 0.7822265625
                },
                {
                    "start": 1364.8,
                    "end": 1364.92,
                    "word": "-word",
                    "probability": 0.6796875
                },
                {
                    "start": 1364.92,
                    "end": 1365.18,
                    "word": " case",
                    "probability": 0.80078125
                },
                {
                    "start": 1365.18,
                    "end": 1365.28,
                    "word": ",",
                    "probability": 0.5703125
                },
                {
                    "start": 1365.28,
                    "end": 1365.32,
                    "word": " it's",
                    "probability": 0.886474609375
                },
                {
                    "start": 1365.42,
                    "end": 1365.52,
                    "word": " a little",
                    "probability": 0.80615234375
                }
            ]
        },
        {
            "id": 219,
            "seek": 136570,
            "start": 1365.52,
            "end": 1371.18,
            "text": " little bit harder to describe here, but basically if you had a sequence of 15 words,",
            "tokens": [
                50860,
                707,
                857,
                6081,
                281,
                6786,
                510,
                11,
                457,
                1936,
                498,
                291,
                632,
                257,
                8310,
                295,
                2119,
                2283,
                11,
                51104
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26285601492169536,
            "compression_ratio": 1.5603864734299517,
            "no_speech_prob": 0.010009765625,
            "words": [
                {
                    "start": 1365.52,
                    "end": 1365.78,
                    "word": " bit",
                    "probability": 0.87744140625
                },
                {
                    "start": 1365.78,
                    "end": 1366.88,
                    "word": " harder",
                    "probability": 0.7021484375
                },
                {
                    "start": 1366.88,
                    "end": 1367.18,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 1367.18,
                    "end": 1367.84,
                    "word": " describe",
                    "probability": 0.76611328125
                },
                {
                    "start": 1367.84,
                    "end": 1368.04,
                    "word": " here",
                    "probability": 0.79541015625
                },
                {
                    "start": 1368.04,
                    "end": 1368.12,
                    "word": ",",
                    "probability": 0.62451171875
                },
                {
                    "start": 1368.12,
                    "end": 1368.18,
                    "word": " but",
                    "probability": 0.89892578125
                },
                {
                    "start": 1368.18,
                    "end": 1368.46,
                    "word": " basically",
                    "probability": 0.74365234375
                },
                {
                    "start": 1368.46,
                    "end": 1368.6,
                    "word": " if",
                    "probability": 0.57080078125
                },
                {
                    "start": 1368.6,
                    "end": 1368.68,
                    "word": " you",
                    "probability": 0.8935546875
                },
                {
                    "start": 1368.68,
                    "end": 1368.82,
                    "word": " had",
                    "probability": 0.5703125
                },
                {
                    "start": 1369.02,
                    "end": 1369.28,
                    "word": " a sequence",
                    "probability": 0.890625
                },
                {
                    "start": 1369.28,
                    "end": 1369.48,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 1369.48,
                    "end": 1369.9,
                    "word": " 15",
                    "probability": 0.693359375
                },
                {
                    "start": 1369.9,
                    "end": 1370.3,
                    "word": " words",
                    "probability": 0.7275390625
                },
                {
                    "start": 1370.3,
                    "end": 1370.98,
                    "word": ",",
                    "probability": 0.8359375
                },
                {
                    "start": 1370.98,
                    "end": 1371.18,
                    "word": " for",
                    "probability": 0.880859375
                }
            ]
        },
        {
            "id": 220,
            "seek": 136570,
            "start": 1371.18,
            "end": 1377.84,
            "text": " for every word, the attention mechanism would compute how much",
            "tokens": [
                51140,
                337,
                633,
                1349,
                11,
                264,
                3202,
                7513,
                576,
                14722,
                577,
                709,
                51372
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26285601492169536,
            "compression_ratio": 1.5603864734299517,
            "no_speech_prob": 0.010009765625,
            "words": [
                {
                    "start": 1371.18,
                    "end": 1371.4,
                    "word": " every",
                    "probability": 0.74951171875
                },
                {
                    "start": 1371.4,
                    "end": 1371.72,
                    "word": " word",
                    "probability": 0.82568359375
                },
                {
                    "start": 1371.72,
                    "end": 1372.14,
                    "word": ",",
                    "probability": 0.7333984375
                },
                {
                    "start": 1372.14,
                    "end": 1372.52,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 1372.52,
                    "end": 1372.8,
                    "word": " attention",
                    "probability": 0.814453125
                },
                {
                    "start": 1372.8,
                    "end": 1373.22,
                    "word": " mechanism",
                    "probability": 0.857421875
                },
                {
                    "start": 1373.22,
                    "end": 1373.62,
                    "word": " would",
                    "probability": 0.84912109375
                },
                {
                    "start": 1373.62,
                    "end": 1374.14,
                    "word": " compute",
                    "probability": 0.8349609375
                },
                {
                    "start": 1374.14,
                    "end": 1375.46,
                    "word": " how",
                    "probability": 0.8876953125
                },
                {
                    "start": 1375.46,
                    "end": 1377.16,
                    "word": " much",
                    "probability": 0.828125
                },
                {
                    "start": 1377.16,
                    "end": 1377.84,
                    "word": " that",
                    "probability": 0.488037109375
                }
            ]
        },
        {
            "id": 221,
            "seek": 136570,
            "start": 1377.84,
            "end": 1384.24,
            "text": " that word pertains to the other 14 words in the sentence. You'll see how that has a",
            "tokens": [
                51424,
                300,
                1349,
                13269,
                2315,
                281,
                264,
                661,
                3499,
                2283,
                294,
                264,
                8174,
                13,
                509,
                603,
                536,
                577,
                300,
                575,
                257,
                51800
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26285601492169536,
            "compression_ratio": 1.5603864734299517,
            "no_speech_prob": 0.010009765625,
            "words": [
                {
                    "start": 1377.84,
                    "end": 1378.22,
                    "word": " word",
                    "probability": 0.80517578125
                },
                {
                    "start": 1378.22,
                    "end": 1378.74,
                    "word": " pertains",
                    "probability": 0.9404296875
                },
                {
                    "start": 1378.74,
                    "end": 1379.04,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1379.04,
                    "end": 1379.22,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1379.22,
                    "end": 1379.4,
                    "word": " other",
                    "probability": 0.79052734375
                },
                {
                    "start": 1379.4,
                    "end": 1379.92,
                    "word": " 14",
                    "probability": 0.9189453125
                },
                {
                    "start": 1379.92,
                    "end": 1380.9,
                    "word": " words",
                    "probability": 0.7275390625
                },
                {
                    "start": 1380.9,
                    "end": 1381.34,
                    "word": " in",
                    "probability": 0.89892578125
                },
                {
                    "start": 1381.34,
                    "end": 1381.48,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 1381.48,
                    "end": 1381.82,
                    "word": " sentence",
                    "probability": 0.85107421875
                },
                {
                    "start": 1381.82,
                    "end": 1382.5,
                    "word": ".",
                    "probability": 0.81689453125
                },
                {
                    "start": 1382.96,
                    "end": 1383.22,
                    "word": " You'll",
                    "probability": 0.699462890625
                },
                {
                    "start": 1383.22,
                    "end": 1383.58,
                    "word": " see",
                    "probability": 0.8154296875
                },
                {
                    "start": 1383.58,
                    "end": 1383.78,
                    "word": " how",
                    "probability": 0.89453125
                },
                {
                    "start": 1383.78,
                    "end": 1383.96,
                    "word": " that",
                    "probability": 0.87939453125
                },
                {
                    "start": 1383.96,
                    "end": 1384.14,
                    "word": " has",
                    "probability": 0.89794921875
                },
                {
                    "start": 1384.14,
                    "end": 1384.24,
                    "word": " a",
                    "probability": 0.98095703125
                }
            ]
        },
        {
            "id": 222,
            "seek": 139570,
            "start": 1384.24,
            "end": 1388.6,
            "text": " relationship with the coreference resolution problem I talked about earlier. But it's just",
            "tokens": [
                50364,
                2480,
                365,
                264,
                4965,
                5158,
                8669,
                1154,
                286,
                2825,
                466,
                3071,
                13,
                583,
                309,
                311,
                445,
                50580
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2268949515642004,
            "compression_ratio": 1.7095435684647302,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1384.24,
                    "end": 1384.62,
                    "word": " relationship",
                    "probability": 0.5859375
                },
                {
                    "start": 1384.62,
                    "end": 1384.84,
                    "word": " with",
                    "probability": 0.80712890625
                },
                {
                    "start": 1384.84,
                    "end": 1385.02,
                    "word": " the",
                    "probability": 0.389404296875
                },
                {
                    "start": 1385.02,
                    "end": 1385.58,
                    "word": " coreference",
                    "probability": 0.660400390625
                },
                {
                    "start": 1385.58,
                    "end": 1386.08,
                    "word": " resolution",
                    "probability": 0.65625
                },
                {
                    "start": 1386.08,
                    "end": 1387.1,
                    "word": " problem",
                    "probability": 0.7275390625
                },
                {
                    "start": 1387.1,
                    "end": 1387.26,
                    "word": " I",
                    "probability": 0.984375
                },
                {
                    "start": 1387.26,
                    "end": 1387.48,
                    "word": " talked",
                    "probability": 0.75732421875
                },
                {
                    "start": 1387.48,
                    "end": 1387.74,
                    "word": " about",
                    "probability": 0.822265625
                },
                {
                    "start": 1387.74,
                    "end": 1388.14,
                    "word": " earlier",
                    "probability": 0.91357421875
                },
                {
                    "start": 1388.14,
                    "end": 1388.28,
                    "word": ".",
                    "probability": 0.7431640625
                },
                {
                    "start": 1388.28,
                    "end": 1388.38,
                    "word": " But",
                    "probability": 0.7509765625
                },
                {
                    "start": 1388.38,
                    "end": 1388.5,
                    "word": " it's",
                    "probability": 0.9248046875
                },
                {
                    "start": 1388.5,
                    "end": 1388.6,
                    "word": " just",
                    "probability": 0.8896484375
                }
            ]
        },
        {
            "id": 223,
            "seek": 139570,
            "start": 1388.6,
            "end": 1394.56,
            "text": " intuitive that there's a graph of word relationships in a sentence, and self-attention",
            "tokens": [
                50580,
                21769,
                300,
                456,
                311,
                257,
                4295,
                295,
                1349,
                6159,
                294,
                257,
                8174,
                11,
                293,
                2698,
                12,
                1591,
                1251,
                50884
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2268949515642004,
            "compression_ratio": 1.7095435684647302,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1388.6,
                    "end": 1388.96,
                    "word": " intuitive",
                    "probability": 0.79833984375
                },
                {
                    "start": 1388.96,
                    "end": 1389.28,
                    "word": " that",
                    "probability": 0.86279296875
                },
                {
                    "start": 1389.28,
                    "end": 1390.28,
                    "word": " there's",
                    "probability": 0.908447265625
                },
                {
                    "start": 1390.4,
                    "end": 1390.72,
                    "word": " a graph",
                    "probability": 0.77001953125
                },
                {
                    "start": 1390.72,
                    "end": 1390.92,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 1390.92,
                    "end": 1391.22,
                    "word": " word",
                    "probability": 0.52685546875
                },
                {
                    "start": 1391.22,
                    "end": 1393.02,
                    "word": " relationships",
                    "probability": 0.65380859375
                },
                {
                    "start": 1393.02,
                    "end": 1393.3,
                    "word": " in",
                    "probability": 0.72509765625
                },
                {
                    "start": 1393.42,
                    "end": 1393.64,
                    "word": " a sentence",
                    "probability": 0.8408203125
                },
                {
                    "start": 1393.64,
                    "end": 1393.88,
                    "word": ",",
                    "probability": 0.306396484375
                },
                {
                    "start": 1393.88,
                    "end": 1393.98,
                    "word": " and",
                    "probability": 0.89111328125
                },
                {
                    "start": 1393.98,
                    "end": 1394.22,
                    "word": " self",
                    "probability": 0.82373046875
                },
                {
                    "start": 1394.22,
                    "end": 1394.56,
                    "word": "-attention",
                    "probability": 0.88916015625
                }
            ]
        },
        {
            "id": 224,
            "seek": 139570,
            "start": 1394.56,
            "end": 1401.82,
            "text": " kind of models that in terms of which subjects pertain to which objects,",
            "tokens": [
                50884,
                733,
                295,
                5245,
                300,
                294,
                2115,
                295,
                597,
                13066,
                13269,
                491,
                281,
                597,
                6565,
                11,
                51188
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2268949515642004,
            "compression_ratio": 1.7095435684647302,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1394.56,
                    "end": 1394.88,
                    "word": " kind",
                    "probability": 0.6298828125
                },
                {
                    "start": 1394.88,
                    "end": 1395.06,
                    "word": " of",
                    "probability": 0.9111328125
                },
                {
                    "start": 1395.06,
                    "end": 1395.68,
                    "word": " models",
                    "probability": 0.794921875
                },
                {
                    "start": 1395.68,
                    "end": 1395.94,
                    "word": " that",
                    "probability": 0.88427734375
                },
                {
                    "start": 1395.94,
                    "end": 1396.1,
                    "word": " in",
                    "probability": 0.73583984375
                },
                {
                    "start": 1396.1,
                    "end": 1396.26,
                    "word": " terms",
                    "probability": 0.7666015625
                },
                {
                    "start": 1396.26,
                    "end": 1396.38,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 1396.38,
                    "end": 1396.54,
                    "word": " which",
                    "probability": 0.89697265625
                },
                {
                    "start": 1396.54,
                    "end": 1396.86,
                    "word": " subjects",
                    "probability": 0.8330078125
                },
                {
                    "start": 1396.86,
                    "end": 1400.06,
                    "word": " pertain",
                    "probability": 0.820068359375
                },
                {
                    "start": 1400.06,
                    "end": 1400.18,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1400.18,
                    "end": 1400.32,
                    "word": " which",
                    "probability": 0.8974609375
                },
                {
                    "start": 1400.32,
                    "end": 1400.8,
                    "word": " objects",
                    "probability": 0.8408203125
                },
                {
                    "start": 1400.8,
                    "end": 1401.82,
                    "word": ",",
                    "probability": 0.93359375
                }
            ]
        },
        {
            "id": 225,
            "seek": 139570,
            "start": 1401.82,
            "end": 1407.76,
            "text": " which pronouns pertain to which people, etc. There's an addition and normalization step",
            "tokens": [
                51240,
                597,
                35883,
                13269,
                491,
                281,
                597,
                561,
                11,
                5183,
                13,
                821,
                311,
                364,
                4500,
                293,
                2710,
                2144,
                1823,
                51544
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2268949515642004,
            "compression_ratio": 1.7095435684647302,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1401.82,
                    "end": 1402.12,
                    "word": " which",
                    "probability": 0.90673828125
                },
                {
                    "start": 1402.12,
                    "end": 1402.62,
                    "word": " pronouns",
                    "probability": 0.7587890625
                },
                {
                    "start": 1402.62,
                    "end": 1403.0,
                    "word": " pertain",
                    "probability": 0.89892578125
                },
                {
                    "start": 1403.0,
                    "end": 1403.14,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 1403.14,
                    "end": 1403.28,
                    "word": " which",
                    "probability": 0.9091796875
                },
                {
                    "start": 1403.28,
                    "end": 1403.54,
                    "word": " people",
                    "probability": 0.9306640625
                },
                {
                    "start": 1403.54,
                    "end": 1403.66,
                    "word": ",",
                    "probability": 0.951171875
                },
                {
                    "start": 1403.66,
                    "end": 1403.78,
                    "word": " etc",
                    "probability": 0.2078857421875
                },
                {
                    "start": 1403.78,
                    "end": 1404.06,
                    "word": ".",
                    "probability": 0.9638671875
                },
                {
                    "start": 1405.7,
                    "end": 1406.22,
                    "word": " There's",
                    "probability": 0.845947265625
                },
                {
                    "start": 1406.22,
                    "end": 1406.38,
                    "word": " an",
                    "probability": 0.89501953125
                },
                {
                    "start": 1406.38,
                    "end": 1406.66,
                    "word": " addition",
                    "probability": 0.86767578125
                },
                {
                    "start": 1406.66,
                    "end": 1406.82,
                    "word": " and",
                    "probability": 0.7451171875
                },
                {
                    "start": 1406.82,
                    "end": 1407.32,
                    "word": " normalization",
                    "probability": 0.938232421875
                },
                {
                    "start": 1407.32,
                    "end": 1407.76,
                    "word": " step",
                    "probability": 0.830078125
                }
            ]
        },
        {
            "id": 226,
            "seek": 139570,
            "start": 1407.76,
            "end": 1413.68,
            "text": " that happens when we've run this self-attention process a number of times",
            "tokens": [
                51544,
                300,
                2314,
                562,
                321,
                600,
                1190,
                341,
                2698,
                12,
                1591,
                1251,
                1399,
                257,
                1230,
                295,
                1413,
                51828
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2268949515642004,
            "compression_ratio": 1.7095435684647302,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 1407.76,
                    "end": 1408.12,
                    "word": " that",
                    "probability": 0.8798828125
                },
                {
                    "start": 1408.12,
                    "end": 1408.66,
                    "word": " happens",
                    "probability": 0.90625
                },
                {
                    "start": 1408.66,
                    "end": 1408.92,
                    "word": " when",
                    "probability": 0.84912109375
                },
                {
                    "start": 1408.92,
                    "end": 1409.6,
                    "word": " we've",
                    "probability": 0.81005859375
                },
                {
                    "start": 1409.6,
                    "end": 1410.42,
                    "word": " run",
                    "probability": 0.51513671875
                },
                {
                    "start": 1410.42,
                    "end": 1410.64,
                    "word": " this",
                    "probability": 0.75
                },
                {
                    "start": 1410.64,
                    "end": 1410.82,
                    "word": " self",
                    "probability": 0.83154296875
                },
                {
                    "start": 1410.82,
                    "end": 1411.22,
                    "word": "-attention",
                    "probability": 0.9454752604166666
                },
                {
                    "start": 1411.22,
                    "end": 1412.36,
                    "word": " process",
                    "probability": 0.8759765625
                },
                {
                    "start": 1412.82,
                    "end": 1413.0,
                    "word": " a number",
                    "probability": 0.87744140625
                },
                {
                    "start": 1413.0,
                    "end": 1413.2,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1413.2,
                    "end": 1413.68,
                    "word": " times",
                    "probability": 0.8310546875
                }
            ]
        },
        {
            "id": 227,
            "seek": 142214,
            "start": 1414.92,
            "end": 1419.46,
            "text": " and typically we don't just do self-attention once. We have a number of what are called heads and",
            "tokens": [
                50376,
                293,
                5850,
                321,
                500,
                380,
                445,
                360,
                2698,
                12,
                1591,
                1251,
                1564,
                13,
                492,
                362,
                257,
                1230,
                295,
                437,
                366,
                1219,
                8050,
                293,
                50632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21239407360553741,
            "compression_ratio": 1.8364312267657992,
            "no_speech_prob": 8.612871170043945e-05,
            "words": [
                {
                    "start": 1414.92,
                    "end": 1415.32,
                    "word": " and",
                    "probability": 0.036895751953125
                },
                {
                    "start": 1415.32,
                    "end": 1415.72,
                    "word": " typically",
                    "probability": 0.658203125
                },
                {
                    "start": 1415.72,
                    "end": 1416.14,
                    "word": " we",
                    "probability": 0.7880859375
                },
                {
                    "start": 1416.14,
                    "end": 1416.3,
                    "word": " don't",
                    "probability": 0.946044921875
                },
                {
                    "start": 1416.3,
                    "end": 1416.46,
                    "word": " just",
                    "probability": 0.88037109375
                },
                {
                    "start": 1416.46,
                    "end": 1416.6,
                    "word": " do",
                    "probability": 0.904296875
                },
                {
                    "start": 1416.6,
                    "end": 1416.84,
                    "word": " self",
                    "probability": 0.8505859375
                },
                {
                    "start": 1416.84,
                    "end": 1417.12,
                    "word": "-attention",
                    "probability": 0.8372395833333334
                },
                {
                    "start": 1417.12,
                    "end": 1417.44,
                    "word": " once",
                    "probability": 0.78759765625
                },
                {
                    "start": 1417.44,
                    "end": 1417.58,
                    "word": ".",
                    "probability": 0.31884765625
                },
                {
                    "start": 1417.58,
                    "end": 1417.64,
                    "word": " We",
                    "probability": 0.88818359375
                },
                {
                    "start": 1417.64,
                    "end": 1417.78,
                    "word": " have",
                    "probability": 0.876953125
                },
                {
                    "start": 1417.86,
                    "end": 1418.02,
                    "word": " a number",
                    "probability": 0.87890625
                },
                {
                    "start": 1418.02,
                    "end": 1418.26,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1418.26,
                    "end": 1418.72,
                    "word": " what",
                    "probability": 0.8173828125
                },
                {
                    "start": 1418.72,
                    "end": 1418.82,
                    "word": " are",
                    "probability": 0.8896484375
                },
                {
                    "start": 1418.82,
                    "end": 1418.98,
                    "word": " called",
                    "probability": 0.8056640625
                },
                {
                    "start": 1418.98,
                    "end": 1419.28,
                    "word": " heads",
                    "probability": 0.7431640625
                },
                {
                    "start": 1419.28,
                    "end": 1419.46,
                    "word": " and",
                    "probability": 0.470458984375
                }
            ]
        },
        {
            "id": 228,
            "seek": 142214,
            "start": 1419.46,
            "end": 1424.52,
            "text": " we run attention a number of times. We combine all that together, we do some non-linear transformation",
            "tokens": [
                50632,
                321,
                1190,
                3202,
                257,
                1230,
                295,
                1413,
                13,
                492,
                10432,
                439,
                300,
                1214,
                11,
                321,
                360,
                512,
                2107,
                12,
                28263,
                9887,
                50888
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21239407360553741,
            "compression_ratio": 1.8364312267657992,
            "no_speech_prob": 8.612871170043945e-05,
            "words": [
                {
                    "start": 1419.46,
                    "end": 1419.88,
                    "word": " we",
                    "probability": 0.82861328125
                },
                {
                    "start": 1419.88,
                    "end": 1420.36,
                    "word": " run",
                    "probability": 0.76708984375
                },
                {
                    "start": 1420.36,
                    "end": 1420.64,
                    "word": " attention",
                    "probability": 0.80224609375
                },
                {
                    "start": 1420.8,
                    "end": 1420.92,
                    "word": " a number",
                    "probability": 0.87890625
                },
                {
                    "start": 1420.92,
                    "end": 1421.02,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 1421.02,
                    "end": 1421.26,
                    "word": " times",
                    "probability": 0.8291015625
                },
                {
                    "start": 1421.26,
                    "end": 1421.66,
                    "word": ".",
                    "probability": 0.90234375
                },
                {
                    "start": 1421.7,
                    "end": 1421.84,
                    "word": " We",
                    "probability": 0.90087890625
                },
                {
                    "start": 1421.84,
                    "end": 1422.12,
                    "word": " combine",
                    "probability": 0.8828125
                },
                {
                    "start": 1422.12,
                    "end": 1422.3,
                    "word": " all",
                    "probability": 0.89599609375
                },
                {
                    "start": 1422.3,
                    "end": 1422.5,
                    "word": " that",
                    "probability": 0.642578125
                },
                {
                    "start": 1422.5,
                    "end": 1422.98,
                    "word": " together",
                    "probability": 0.7138671875
                },
                {
                    "start": 1422.98,
                    "end": 1423.2,
                    "word": ",",
                    "probability": 0.57177734375
                },
                {
                    "start": 1423.2,
                    "end": 1423.38,
                    "word": " we",
                    "probability": 0.8828125
                },
                {
                    "start": 1423.38,
                    "end": 1423.52,
                    "word": " do",
                    "probability": 0.90380859375
                },
                {
                    "start": 1423.52,
                    "end": 1423.7,
                    "word": " some",
                    "probability": 0.806640625
                },
                {
                    "start": 1423.7,
                    "end": 1423.9,
                    "word": " non",
                    "probability": 0.9287109375
                },
                {
                    "start": 1423.9,
                    "end": 1424.06,
                    "word": "-linear",
                    "probability": 0.673828125
                },
                {
                    "start": 1424.06,
                    "end": 1424.52,
                    "word": " transformation",
                    "probability": 0.77099609375
                }
            ]
        },
        {
            "id": 229,
            "seek": 142214,
            "start": 1424.52,
            "end": 1429.8,
            "text": " on it, and then we wind up with a new embedding of your original sequence of the same dimensionality",
            "tokens": [
                50888,
                322,
                309,
                11,
                293,
                550,
                321,
                2468,
                493,
                365,
                257,
                777,
                12240,
                3584,
                295,
                428,
                3380,
                8310,
                295,
                264,
                912,
                10139,
                1860,
                51140
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21239407360553741,
            "compression_ratio": 1.8364312267657992,
            "no_speech_prob": 8.612871170043945e-05,
            "words": [
                {
                    "start": 1424.52,
                    "end": 1424.9,
                    "word": " on",
                    "probability": 0.90576171875
                },
                {
                    "start": 1424.9,
                    "end": 1425.02,
                    "word": " it",
                    "probability": 0.904296875
                },
                {
                    "start": 1425.02,
                    "end": 1425.04,
                    "word": ",",
                    "probability": 0.56494140625
                },
                {
                    "start": 1425.04,
                    "end": 1425.1,
                    "word": " and",
                    "probability": 0.8837890625
                },
                {
                    "start": 1425.1,
                    "end": 1425.2,
                    "word": " then",
                    "probability": 0.71728515625
                },
                {
                    "start": 1425.2,
                    "end": 1425.3,
                    "word": " we",
                    "probability": 0.91162109375
                },
                {
                    "start": 1425.3,
                    "end": 1425.48,
                    "word": " wind",
                    "probability": 0.86279296875
                },
                {
                    "start": 1425.48,
                    "end": 1425.6,
                    "word": " up",
                    "probability": 0.90869140625
                },
                {
                    "start": 1425.6,
                    "end": 1425.72,
                    "word": " with",
                    "probability": 0.81494140625
                },
                {
                    "start": 1425.9,
                    "end": 1426.28,
                    "word": " a new",
                    "probability": 0.82763671875
                },
                {
                    "start": 1426.28,
                    "end": 1427.48,
                    "word": " embedding",
                    "probability": 0.8662109375
                },
                {
                    "start": 1427.48,
                    "end": 1427.72,
                    "word": " of",
                    "probability": 0.87744140625
                },
                {
                    "start": 1427.72,
                    "end": 1427.88,
                    "word": " your",
                    "probability": 0.446044921875
                },
                {
                    "start": 1427.88,
                    "end": 1428.5,
                    "word": " original",
                    "probability": 0.8291015625
                },
                {
                    "start": 1428.5,
                    "end": 1428.86,
                    "word": " sequence",
                    "probability": 0.916015625
                },
                {
                    "start": 1428.86,
                    "end": 1428.98,
                    "word": " of",
                    "probability": 0.642578125
                },
                {
                    "start": 1428.98,
                    "end": 1429.08,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1429.08,
                    "end": 1429.24,
                    "word": " same",
                    "probability": 0.79150390625
                },
                {
                    "start": 1429.24,
                    "end": 1429.8,
                    "word": " dimensionality",
                    "probability": 0.870849609375
                }
            ]
        },
        {
            "id": 230,
            "seek": 142214,
            "start": 1429.8,
            "end": 1436.0,
            "text": " as this original embedding, except that now thinking is encoded in this new representation",
            "tokens": [
                51140,
                382,
                341,
                3380,
                12240,
                3584,
                11,
                3993,
                300,
                586,
                1953,
                307,
                2058,
                12340,
                294,
                341,
                777,
                10290,
                51464
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21239407360553741,
            "compression_ratio": 1.8364312267657992,
            "no_speech_prob": 8.612871170043945e-05,
            "words": [
                {
                    "start": 1429.8,
                    "end": 1429.94,
                    "word": " as",
                    "probability": 0.88818359375
                },
                {
                    "start": 1429.94,
                    "end": 1430.08,
                    "word": " this",
                    "probability": 0.8173828125
                },
                {
                    "start": 1430.08,
                    "end": 1430.42,
                    "word": " original",
                    "probability": 0.83154296875
                },
                {
                    "start": 1430.42,
                    "end": 1431.16,
                    "word": " embedding",
                    "probability": 0.866455078125
                },
                {
                    "start": 1431.16,
                    "end": 1431.18,
                    "word": ",",
                    "probability": 0.5625
                },
                {
                    "start": 1431.18,
                    "end": 1431.4,
                    "word": " except",
                    "probability": 0.876953125
                },
                {
                    "start": 1431.4,
                    "end": 1431.6,
                    "word": " that",
                    "probability": 0.87548828125
                },
                {
                    "start": 1431.6,
                    "end": 1431.78,
                    "word": " now",
                    "probability": 0.8759765625
                },
                {
                    "start": 1431.78,
                    "end": 1432.16,
                    "word": " thinking",
                    "probability": 0.58544921875
                },
                {
                    "start": 1432.16,
                    "end": 1432.88,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 1432.88,
                    "end": 1433.84,
                    "word": " encoded",
                    "probability": 0.880615234375
                },
                {
                    "start": 1433.84,
                    "end": 1434.5,
                    "word": " in",
                    "probability": 0.8955078125
                },
                {
                    "start": 1434.5,
                    "end": 1435.42,
                    "word": " this",
                    "probability": 0.7939453125
                },
                {
                    "start": 1435.42,
                    "end": 1435.6,
                    "word": " new",
                    "probability": 0.83935546875
                },
                {
                    "start": 1435.6,
                    "end": 1436.0,
                    "word": " representation",
                    "probability": 0.78173828125
                }
            ]
        },
        {
            "id": 231,
            "seek": 142214,
            "start": 1436.0,
            "end": 1440.68,
            "text": " in the context in which it appears, in the context of machines. Machines is encoded in the context of",
            "tokens": [
                51464,
                294,
                264,
                4319,
                294,
                597,
                309,
                7038,
                11,
                294,
                264,
                4319,
                295,
                8379,
                13,
                12089,
                1652,
                307,
                2058,
                12340,
                294,
                264,
                4319,
                295,
                51684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21239407360553741,
            "compression_ratio": 1.8364312267657992,
            "no_speech_prob": 8.612871170043945e-05,
            "words": [
                {
                    "start": 1436.0,
                    "end": 1436.4,
                    "word": " in",
                    "probability": 0.6826171875
                },
                {
                    "start": 1436.4,
                    "end": 1436.56,
                    "word": " the",
                    "probability": 0.82080078125
                },
                {
                    "start": 1436.56,
                    "end": 1436.92,
                    "word": " context",
                    "probability": 0.755859375
                },
                {
                    "start": 1436.92,
                    "end": 1437.12,
                    "word": " in",
                    "probability": 0.90185546875
                },
                {
                    "start": 1437.12,
                    "end": 1437.22,
                    "word": " which",
                    "probability": 0.91162109375
                },
                {
                    "start": 1437.22,
                    "end": 1437.32,
                    "word": " it",
                    "probability": 0.9033203125
                },
                {
                    "start": 1437.32,
                    "end": 1437.68,
                    "word": " appears",
                    "probability": 0.75048828125
                },
                {
                    "start": 1437.68,
                    "end": 1437.78,
                    "word": ",",
                    "probability": 0.46826171875
                },
                {
                    "start": 1437.78,
                    "end": 1437.84,
                    "word": " in",
                    "probability": 0.89111328125
                },
                {
                    "start": 1437.84,
                    "end": 1437.94,
                    "word": " the",
                    "probability": 0.818359375
                },
                {
                    "start": 1437.94,
                    "end": 1438.18,
                    "word": " context",
                    "probability": 0.75439453125
                },
                {
                    "start": 1438.18,
                    "end": 1438.38,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 1438.38,
                    "end": 1438.84,
                    "word": " machines",
                    "probability": 0.8037109375
                },
                {
                    "start": 1438.84,
                    "end": 1439.04,
                    "word": ".",
                    "probability": 0.79052734375
                },
                {
                    "start": 1439.04,
                    "end": 1439.42,
                    "word": " Machines",
                    "probability": 0.923583984375
                },
                {
                    "start": 1439.42,
                    "end": 1439.6,
                    "word": " is",
                    "probability": 0.84423828125
                },
                {
                    "start": 1439.6,
                    "end": 1439.9,
                    "word": " encoded",
                    "probability": 0.939697265625
                },
                {
                    "start": 1439.9,
                    "end": 1440.04,
                    "word": " in",
                    "probability": 0.90283203125
                },
                {
                    "start": 1440.04,
                    "end": 1440.16,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 1440.16,
                    "end": 1440.46,
                    "word": " context",
                    "probability": 0.755859375
                },
                {
                    "start": 1440.46,
                    "end": 1440.68,
                    "word": " of",
                    "probability": 0.90673828125
                }
            ]
        },
        {
            "id": 232,
            "seek": 144744,
            "start": 1440.68,
            "end": 1444.4,
            "text": " thinking. Then typically we stack these transformer blocks that we actually do.",
            "tokens": [
                50364,
                1953,
                13,
                1396,
                5850,
                321,
                8630,
                613,
                31782,
                8474,
                300,
                321,
                767,
                360,
                13,
                50552
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26728724450507063,
            "compression_ratio": 1.7958333333333334,
            "no_speech_prob": 8.094310760498047e-05,
            "words": [
                {
                    "start": 1440.68,
                    "end": 1440.94,
                    "word": " thinking",
                    "probability": 0.1982421875
                },
                {
                    "start": 1440.94,
                    "end": 1441.42,
                    "word": ".",
                    "probability": 0.1932373046875
                },
                {
                    "start": 1441.62,
                    "end": 1441.74,
                    "word": " Then",
                    "probability": 0.35546875
                },
                {
                    "start": 1441.74,
                    "end": 1441.94,
                    "word": " typically",
                    "probability": 0.74658203125
                },
                {
                    "start": 1441.94,
                    "end": 1442.12,
                    "word": " we",
                    "probability": 0.767578125
                },
                {
                    "start": 1442.12,
                    "end": 1442.44,
                    "word": " stack",
                    "probability": 0.86083984375
                },
                {
                    "start": 1442.44,
                    "end": 1442.66,
                    "word": " these",
                    "probability": 0.7158203125
                },
                {
                    "start": 1442.66,
                    "end": 1443.02,
                    "word": " transformer",
                    "probability": 0.7900390625
                },
                {
                    "start": 1443.02,
                    "end": 1443.36,
                    "word": " blocks",
                    "probability": 0.80078125
                },
                {
                    "start": 1443.36,
                    "end": 1443.48,
                    "word": " that",
                    "probability": 0.207763671875
                },
                {
                    "start": 1443.48,
                    "end": 1443.56,
                    "word": " we",
                    "probability": 0.89599609375
                },
                {
                    "start": 1443.56,
                    "end": 1443.74,
                    "word": " actually",
                    "probability": 0.7783203125
                },
                {
                    "start": 1443.74,
                    "end": 1444.04,
                    "word": " do",
                    "probability": 0.89013671875
                },
                {
                    "start": 1444.04,
                    "end": 1444.4,
                    "word": ".",
                    "probability": 0.58251953125
                }
            ]
        },
        {
            "id": 233,
            "seek": 144744,
            "start": 1444.4,
            "end": 1447.84,
            "text": " We have another transformer block that then sort of refines the representation and we keep",
            "tokens": [
                50552,
                492,
                362,
                1071,
                31782,
                3461,
                300,
                550,
                1333,
                295,
                1895,
                1652,
                264,
                10290,
                293,
                321,
                1066,
                50736
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26728724450507063,
            "compression_ratio": 1.7958333333333334,
            "no_speech_prob": 8.094310760498047e-05,
            "words": [
                {
                    "start": 1444.4,
                    "end": 1444.64,
                    "word": " We",
                    "probability": 0.89892578125
                },
                {
                    "start": 1444.64,
                    "end": 1444.74,
                    "word": " have",
                    "probability": 0.88037109375
                },
                {
                    "start": 1444.74,
                    "end": 1444.94,
                    "word": " another",
                    "probability": 0.8310546875
                },
                {
                    "start": 1444.94,
                    "end": 1445.3,
                    "word": " transformer",
                    "probability": 0.814453125
                },
                {
                    "start": 1445.3,
                    "end": 1445.58,
                    "word": " block",
                    "probability": 0.7841796875
                },
                {
                    "start": 1445.58,
                    "end": 1445.74,
                    "word": " that",
                    "probability": 0.8603515625
                },
                {
                    "start": 1445.74,
                    "end": 1446.02,
                    "word": " then",
                    "probability": 0.69677734375
                },
                {
                    "start": 1446.02,
                    "end": 1446.28,
                    "word": " sort",
                    "probability": 0.77099609375
                },
                {
                    "start": 1446.28,
                    "end": 1446.34,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1446.34,
                    "end": 1446.68,
                    "word": " refines",
                    "probability": 0.823974609375
                },
                {
                    "start": 1446.68,
                    "end": 1446.82,
                    "word": " the",
                    "probability": 0.5693359375
                },
                {
                    "start": 1446.82,
                    "end": 1447.28,
                    "word": " representation",
                    "probability": 0.728515625
                },
                {
                    "start": 1447.28,
                    "end": 1447.52,
                    "word": " and",
                    "probability": 0.60498046875
                },
                {
                    "start": 1447.52,
                    "end": 1447.64,
                    "word": " we",
                    "probability": 0.896484375
                },
                {
                    "start": 1447.64,
                    "end": 1447.84,
                    "word": " keep",
                    "probability": 0.83056640625
                }
            ]
        },
        {
            "id": 234,
            "seek": 144744,
            "start": 1447.84,
            "end": 1452.5,
            "text": " going. Young will show that we use a number of transformer blocks in our phishing detection",
            "tokens": [
                50736,
                516,
                13,
                8160,
                486,
                855,
                300,
                321,
                764,
                257,
                1230,
                295,
                31782,
                8474,
                294,
                527,
                903,
                3807,
                17784,
                50968
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26728724450507063,
            "compression_ratio": 1.7958333333333334,
            "no_speech_prob": 8.094310760498047e-05,
            "words": [
                {
                    "start": 1447.84,
                    "end": 1448.44,
                    "word": " going",
                    "probability": 0.515625
                },
                {
                    "start": 1448.44,
                    "end": 1448.92,
                    "word": ".",
                    "probability": 0.92431640625
                },
                {
                    "start": 1449.4,
                    "end": 1449.62,
                    "word": " Young",
                    "probability": 0.291259765625
                },
                {
                    "start": 1449.62,
                    "end": 1449.74,
                    "word": " will",
                    "probability": 0.59375
                },
                {
                    "start": 1449.74,
                    "end": 1449.94,
                    "word": " show",
                    "probability": 0.86181640625
                },
                {
                    "start": 1449.94,
                    "end": 1450.06,
                    "word": " that",
                    "probability": 0.85400390625
                },
                {
                    "start": 1450.06,
                    "end": 1450.16,
                    "word": " we",
                    "probability": 0.904296875
                },
                {
                    "start": 1450.16,
                    "end": 1450.42,
                    "word": " use",
                    "probability": 0.751953125
                },
                {
                    "start": 1450.62,
                    "end": 1450.82,
                    "word": " a number",
                    "probability": 0.884765625
                },
                {
                    "start": 1450.82,
                    "end": 1450.96,
                    "word": " of",
                    "probability": 0.90869140625
                },
                {
                    "start": 1450.96,
                    "end": 1451.3,
                    "word": " transformer",
                    "probability": 0.80322265625
                },
                {
                    "start": 1451.3,
                    "end": 1451.66,
                    "word": " blocks",
                    "probability": 0.814453125
                },
                {
                    "start": 1451.66,
                    "end": 1451.78,
                    "word": " in",
                    "probability": 0.8212890625
                },
                {
                    "start": 1451.78,
                    "end": 1451.9,
                    "word": " our",
                    "probability": 0.7802734375
                },
                {
                    "start": 1451.9,
                    "end": 1452.14,
                    "word": " phishing",
                    "probability": 0.73095703125
                },
                {
                    "start": 1452.14,
                    "end": 1452.5,
                    "word": " detection",
                    "probability": 0.7958984375
                }
            ]
        },
        {
            "id": 235,
            "seek": 144744,
            "start": 1452.5,
            "end": 1459.24,
            "text": " work in a few minutes. Here's some intuition as to what comes out of the attention mechanism",
            "tokens": [
                50968,
                589,
                294,
                257,
                1326,
                2077,
                13,
                1692,
                311,
                512,
                24002,
                382,
                281,
                437,
                1487,
                484,
                295,
                264,
                3202,
                7513,
                51292
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26728724450507063,
            "compression_ratio": 1.7958333333333334,
            "no_speech_prob": 8.094310760498047e-05,
            "words": [
                {
                    "start": 1452.5,
                    "end": 1453.06,
                    "word": " work",
                    "probability": 0.78955078125
                },
                {
                    "start": 1453.06,
                    "end": 1453.2,
                    "word": " in",
                    "probability": 0.89599609375
                },
                {
                    "start": 1453.3,
                    "end": 1453.4,
                    "word": " a few",
                    "probability": 0.80517578125
                },
                {
                    "start": 1453.4,
                    "end": 1453.72,
                    "word": " minutes",
                    "probability": 0.88623046875
                },
                {
                    "start": 1453.72,
                    "end": 1454.2,
                    "word": ".",
                    "probability": 0.9833984375
                },
                {
                    "start": 1455.72,
                    "end": 1456.2,
                    "word": " Here's",
                    "probability": 0.827880859375
                },
                {
                    "start": 1456.2,
                    "end": 1456.3,
                    "word": " some",
                    "probability": 0.7666015625
                },
                {
                    "start": 1456.3,
                    "end": 1456.66,
                    "word": " intuition",
                    "probability": 0.8037109375
                },
                {
                    "start": 1456.66,
                    "end": 1457.3,
                    "word": " as",
                    "probability": 0.70068359375
                },
                {
                    "start": 1457.3,
                    "end": 1457.44,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 1457.44,
                    "end": 1457.62,
                    "word": " what",
                    "probability": 0.884765625
                },
                {
                    "start": 1457.62,
                    "end": 1457.98,
                    "word": " comes",
                    "probability": 0.80419921875
                },
                {
                    "start": 1457.98,
                    "end": 1458.16,
                    "word": " out",
                    "probability": 0.8251953125
                },
                {
                    "start": 1458.16,
                    "end": 1458.3,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1458.3,
                    "end": 1458.44,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 1458.44,
                    "end": 1458.72,
                    "word": " attention",
                    "probability": 0.6357421875
                },
                {
                    "start": 1458.72,
                    "end": 1459.24,
                    "word": " mechanism",
                    "probability": 0.89697265625
                }
            ]
        },
        {
            "id": 236,
            "seek": 144744,
            "start": 1460.22,
            "end": 1465.98,
            "text": " in a typical transformer. So here's what a transformer block has decided it",
            "tokens": [
                51348,
                294,
                257,
                7476,
                31782,
                13,
                407,
                510,
                311,
                437,
                257,
                31782,
                3461,
                575,
                3047,
                309,
                51620
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26728724450507063,
            "compression_ratio": 1.7958333333333334,
            "no_speech_prob": 8.094310760498047e-05,
            "words": [
                {
                    "start": 1460.22,
                    "end": 1460.7,
                    "word": " in",
                    "probability": 0.76220703125
                },
                {
                    "start": 1461.38,
                    "end": 1461.66,
                    "word": " a typical",
                    "probability": 0.7841796875
                },
                {
                    "start": 1461.66,
                    "end": 1462.18,
                    "word": " transformer",
                    "probability": 0.810546875
                },
                {
                    "start": 1462.18,
                    "end": 1462.58,
                    "word": ".",
                    "probability": 0.97509765625
                },
                {
                    "start": 1462.58,
                    "end": 1462.68,
                    "word": " So",
                    "probability": 0.82763671875
                },
                {
                    "start": 1462.68,
                    "end": 1463.7,
                    "word": " here's",
                    "probability": 0.838623046875
                },
                {
                    "start": 1463.7,
                    "end": 1463.9,
                    "word": " what",
                    "probability": 0.79443359375
                },
                {
                    "start": 1464.12,
                    "end": 1464.44,
                    "word": " a transformer",
                    "probability": 0.8076171875
                },
                {
                    "start": 1464.44,
                    "end": 1464.84,
                    "word": " block",
                    "probability": 0.81640625
                },
                {
                    "start": 1464.84,
                    "end": 1465.04,
                    "word": " has",
                    "probability": 0.8916015625
                },
                {
                    "start": 1465.04,
                    "end": 1465.44,
                    "word": " decided",
                    "probability": 0.88671875
                },
                {
                    "start": 1465.44,
                    "end": 1465.98,
                    "word": " it",
                    "probability": 0.8701171875
                }
            ]
        },
        {
            "id": 237,
            "seek": 147566,
            "start": 1467.72,
            "end": 1471.44,
            "text": " sort of relates to in an input sentence. So here we have the input sentence,",
            "tokens": [
                50380,
                1333,
                295,
                16155,
                281,
                294,
                364,
                4846,
                8174,
                13,
                407,
                510,
                321,
                362,
                264,
                4846,
                8174,
                11,
                50632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2501716980567345,
            "compression_ratio": 1.7955555555555556,
            "no_speech_prob": 0.00023412704467773438,
            "words": [
                {
                    "start": 1467.72,
                    "end": 1468.24,
                    "word": " sort",
                    "probability": 0.237060546875
                },
                {
                    "start": 1468.24,
                    "end": 1468.4,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 1468.4,
                    "end": 1468.76,
                    "word": " relates",
                    "probability": 0.8671875
                },
                {
                    "start": 1468.76,
                    "end": 1469.04,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 1469.04,
                    "end": 1469.24,
                    "word": " in",
                    "probability": 0.333251953125
                },
                {
                    "start": 1469.24,
                    "end": 1469.32,
                    "word": " an",
                    "probability": 0.5517578125
                },
                {
                    "start": 1469.32,
                    "end": 1469.58,
                    "word": " input",
                    "probability": 0.85498046875
                },
                {
                    "start": 1469.58,
                    "end": 1469.9,
                    "word": " sentence",
                    "probability": 0.83056640625
                },
                {
                    "start": 1469.9,
                    "end": 1470.02,
                    "word": ".",
                    "probability": 0.880859375
                },
                {
                    "start": 1470.02,
                    "end": 1470.1,
                    "word": " So",
                    "probability": 0.83837890625
                },
                {
                    "start": 1470.1,
                    "end": 1470.24,
                    "word": " here",
                    "probability": 0.744140625
                },
                {
                    "start": 1470.24,
                    "end": 1470.34,
                    "word": " we",
                    "probability": 0.8564453125
                },
                {
                    "start": 1470.34,
                    "end": 1470.5,
                    "word": " have",
                    "probability": 0.8720703125
                },
                {
                    "start": 1470.5,
                    "end": 1470.82,
                    "word": " the",
                    "probability": 0.61279296875
                },
                {
                    "start": 1470.82,
                    "end": 1471.06,
                    "word": " input",
                    "probability": 0.84765625
                },
                {
                    "start": 1471.06,
                    "end": 1471.36,
                    "word": " sentence",
                    "probability": 0.87109375
                },
                {
                    "start": 1471.36,
                    "end": 1471.44,
                    "word": ",",
                    "probability": 0.5185546875
                }
            ]
        },
        {
            "id": 238,
            "seek": 147566,
            "start": 1471.44,
            "end": 1474.88,
            "text": " the animal didn't cross the street because it was too tired.",
            "tokens": [
                50632,
                264,
                5496,
                994,
                380,
                3278,
                264,
                4838,
                570,
                309,
                390,
                886,
                5868,
                13,
                50796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2501716980567345,
            "compression_ratio": 1.7955555555555556,
            "no_speech_prob": 0.00023412704467773438,
            "words": [
                {
                    "start": 1471.44,
                    "end": 1471.52,
                    "word": " the",
                    "probability": 0.454833984375
                },
                {
                    "start": 1471.52,
                    "end": 1471.86,
                    "word": " animal",
                    "probability": 0.69775390625
                },
                {
                    "start": 1471.86,
                    "end": 1472.54,
                    "word": " didn't",
                    "probability": 0.9228515625
                },
                {
                    "start": 1472.54,
                    "end": 1472.94,
                    "word": " cross",
                    "probability": 0.828125
                },
                {
                    "start": 1472.94,
                    "end": 1473.14,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1473.14,
                    "end": 1473.4,
                    "word": " street",
                    "probability": 0.787109375
                },
                {
                    "start": 1473.4,
                    "end": 1473.72,
                    "word": " because",
                    "probability": 0.80224609375
                },
                {
                    "start": 1473.72,
                    "end": 1473.88,
                    "word": " it",
                    "probability": 0.89794921875
                },
                {
                    "start": 1473.88,
                    "end": 1474.04,
                    "word": " was",
                    "probability": 0.8994140625
                },
                {
                    "start": 1474.04,
                    "end": 1474.24,
                    "word": " too",
                    "probability": 0.82373046875
                },
                {
                    "start": 1474.24,
                    "end": 1474.62,
                    "word": " tired",
                    "probability": 0.69189453125
                },
                {
                    "start": 1474.62,
                    "end": 1474.88,
                    "word": ".",
                    "probability": 0.84033203125
                }
            ]
        },
        {
            "id": 239,
            "seek": 147566,
            "start": 1476.44,
            "end": 1483.7,
            "text": " And the strongest attentional relationship here goes to the animal, which is interesting",
            "tokens": [
                50884,
                400,
                264,
                16595,
                3202,
                304,
                2480,
                510,
                1709,
                281,
                264,
                5496,
                11,
                597,
                307,
                1880,
                51256
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2501716980567345,
            "compression_ratio": 1.7955555555555556,
            "no_speech_prob": 0.00023412704467773438,
            "words": [
                {
                    "start": 1476.44,
                    "end": 1476.72,
                    "word": " And",
                    "probability": 0.861328125
                },
                {
                    "start": 1476.72,
                    "end": 1476.9,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 1476.9,
                    "end": 1477.58,
                    "word": " strongest",
                    "probability": 0.84130859375
                },
                {
                    "start": 1477.58,
                    "end": 1478.28,
                    "word": " attentional",
                    "probability": 0.84912109375
                },
                {
                    "start": 1478.28,
                    "end": 1478.7,
                    "word": " relationship",
                    "probability": 0.83984375
                },
                {
                    "start": 1478.7,
                    "end": 1479.22,
                    "word": " here",
                    "probability": 0.80810546875
                },
                {
                    "start": 1479.22,
                    "end": 1481.18,
                    "word": " goes",
                    "probability": 0.7587890625
                },
                {
                    "start": 1481.18,
                    "end": 1481.38,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 1481.38,
                    "end": 1481.54,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 1481.54,
                    "end": 1481.86,
                    "word": " animal",
                    "probability": 0.71484375
                },
                {
                    "start": 1481.86,
                    "end": 1482.24,
                    "word": ",",
                    "probability": 0.6875
                },
                {
                    "start": 1482.24,
                    "end": 1483.18,
                    "word": " which",
                    "probability": 0.91943359375
                },
                {
                    "start": 1483.18,
                    "end": 1483.32,
                    "word": " is",
                    "probability": 0.9013671875
                },
                {
                    "start": 1483.32,
                    "end": 1483.7,
                    "word": " interesting",
                    "probability": 0.6396484375
                }
            ]
        },
        {
            "id": 240,
            "seek": 147566,
            "start": 1483.7,
            "end": 1489.14,
            "text": " because one could ask whether or not it refers to the street or the animal here.",
            "tokens": [
                51256,
                570,
                472,
                727,
                1029,
                1968,
                420,
                406,
                309,
                14942,
                281,
                264,
                4838,
                420,
                264,
                5496,
                510,
                13,
                51492
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2501716980567345,
            "compression_ratio": 1.7955555555555556,
            "no_speech_prob": 0.00023412704467773438,
            "words": [
                {
                    "start": 1483.7,
                    "end": 1484.28,
                    "word": " because",
                    "probability": 0.6826171875
                },
                {
                    "start": 1484.28,
                    "end": 1485.12,
                    "word": " one",
                    "probability": 0.806640625
                },
                {
                    "start": 1485.12,
                    "end": 1485.28,
                    "word": " could",
                    "probability": 0.79541015625
                },
                {
                    "start": 1485.28,
                    "end": 1485.56,
                    "word": " ask",
                    "probability": 0.810546875
                },
                {
                    "start": 1485.56,
                    "end": 1485.76,
                    "word": " whether",
                    "probability": 0.85009765625
                },
                {
                    "start": 1485.76,
                    "end": 1485.96,
                    "word": " or",
                    "probability": 0.90087890625
                },
                {
                    "start": 1485.96,
                    "end": 1486.1,
                    "word": " not",
                    "probability": 0.8955078125
                },
                {
                    "start": 1486.1,
                    "end": 1486.3,
                    "word": " it",
                    "probability": 0.9033203125
                },
                {
                    "start": 1486.3,
                    "end": 1486.58,
                    "word": " refers",
                    "probability": 0.78564453125
                },
                {
                    "start": 1486.58,
                    "end": 1486.76,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1486.76,
                    "end": 1486.94,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 1486.94,
                    "end": 1487.38,
                    "word": " street",
                    "probability": 0.66748046875
                },
                {
                    "start": 1487.38,
                    "end": 1487.78,
                    "word": " or",
                    "probability": 0.8818359375
                },
                {
                    "start": 1487.78,
                    "end": 1487.96,
                    "word": " the",
                    "probability": 0.8193359375
                },
                {
                    "start": 1487.96,
                    "end": 1488.24,
                    "word": " animal",
                    "probability": 0.71875
                },
                {
                    "start": 1488.24,
                    "end": 1488.62,
                    "word": " here",
                    "probability": 0.81201171875
                },
                {
                    "start": 1488.62,
                    "end": 1489.14,
                    "word": ".",
                    "probability": 0.85791015625
                }
            ]
        },
        {
            "id": 241,
            "seek": 147566,
            "start": 1489.78,
            "end": 1494.2,
            "text": " One can interpret the weight of the connection to the animal, meaning that the transformer block",
            "tokens": [
                51564,
                1485,
                393,
                7302,
                264,
                3364,
                295,
                264,
                4984,
                281,
                264,
                5496,
                11,
                3620,
                300,
                264,
                31782,
                3461,
                51768
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2501716980567345,
            "compression_ratio": 1.7955555555555556,
            "no_speech_prob": 0.00023412704467773438,
            "words": [
                {
                    "start": 1489.78,
                    "end": 1490.3,
                    "word": " One",
                    "probability": 0.68115234375
                },
                {
                    "start": 1490.3,
                    "end": 1490.44,
                    "word": " can",
                    "probability": 0.80419921875
                },
                {
                    "start": 1490.44,
                    "end": 1490.78,
                    "word": " interpret",
                    "probability": 0.78515625
                },
                {
                    "start": 1490.78,
                    "end": 1491.26,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 1491.26,
                    "end": 1491.48,
                    "word": " weight",
                    "probability": 0.8505859375
                },
                {
                    "start": 1491.48,
                    "end": 1491.64,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 1491.64,
                    "end": 1491.74,
                    "word": " the",
                    "probability": 0.81982421875
                },
                {
                    "start": 1491.74,
                    "end": 1492.0,
                    "word": " connection",
                    "probability": 0.74267578125
                },
                {
                    "start": 1492.0,
                    "end": 1492.14,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1492.14,
                    "end": 1492.28,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 1492.28,
                    "end": 1492.78,
                    "word": " animal",
                    "probability": 0.72509765625
                },
                {
                    "start": 1492.78,
                    "end": 1492.92,
                    "word": ",",
                    "probability": 0.7607421875
                },
                {
                    "start": 1492.92,
                    "end": 1493.08,
                    "word": " meaning",
                    "probability": 0.80126953125
                },
                {
                    "start": 1493.08,
                    "end": 1493.26,
                    "word": " that",
                    "probability": 0.8701171875
                },
                {
                    "start": 1493.26,
                    "end": 1493.4,
                    "word": " the",
                    "probability": 0.8212890625
                },
                {
                    "start": 1493.4,
                    "end": 1493.78,
                    "word": " transformer",
                    "probability": 0.767578125
                },
                {
                    "start": 1493.78,
                    "end": 1494.2,
                    "word": " block",
                    "probability": 0.78271484375
                }
            ]
        },
        {
            "id": 242,
            "seek": 150350,
            "start": 1494.96,
            "end": 1500.76,
            "text": " has decided, in their quotes, that it pertains to the animal, which is really interesting.",
            "tokens": [
                50408,
                575,
                3047,
                11,
                294,
                641,
                19963,
                11,
                300,
                309,
                13269,
                2315,
                281,
                264,
                5496,
                11,
                597,
                307,
                534,
                1880,
                13,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22926979725903804,
            "compression_ratio": 1.6765799256505576,
            "no_speech_prob": 0.00037980079650878906,
            "words": [
                {
                    "start": 1494.96,
                    "end": 1495.5,
                    "word": " has",
                    "probability": 0.38720703125
                },
                {
                    "start": 1495.5,
                    "end": 1496.04,
                    "word": " decided",
                    "probability": 0.7001953125
                },
                {
                    "start": 1496.04,
                    "end": 1496.36,
                    "word": ",",
                    "probability": 0.55419921875
                },
                {
                    "start": 1496.36,
                    "end": 1496.44,
                    "word": " in",
                    "probability": 0.79052734375
                },
                {
                    "start": 1496.44,
                    "end": 1496.76,
                    "word": " their",
                    "probability": 0.79931640625
                },
                {
                    "start": 1496.76,
                    "end": 1497.1,
                    "word": " quotes",
                    "probability": 0.671875
                },
                {
                    "start": 1497.1,
                    "end": 1497.34,
                    "word": ",",
                    "probability": 0.93505859375
                },
                {
                    "start": 1497.34,
                    "end": 1497.92,
                    "word": " that",
                    "probability": 0.8427734375
                },
                {
                    "start": 1497.92,
                    "end": 1498.32,
                    "word": " it",
                    "probability": 0.861328125
                },
                {
                    "start": 1498.32,
                    "end": 1498.64,
                    "word": " pertains",
                    "probability": 0.9453125
                },
                {
                    "start": 1498.64,
                    "end": 1498.8,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 1498.8,
                    "end": 1498.94,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 1498.94,
                    "end": 1499.2,
                    "word": " animal",
                    "probability": 0.73046875
                },
                {
                    "start": 1499.2,
                    "end": 1499.36,
                    "word": ",",
                    "probability": 0.771484375
                },
                {
                    "start": 1499.36,
                    "end": 1499.48,
                    "word": " which",
                    "probability": 0.9140625
                },
                {
                    "start": 1499.48,
                    "end": 1499.58,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 1499.58,
                    "end": 1499.78,
                    "word": " really",
                    "probability": 0.76806640625
                },
                {
                    "start": 1499.78,
                    "end": 1500.22,
                    "word": " interesting",
                    "probability": 0.64404296875
                },
                {
                    "start": 1500.22,
                    "end": 1500.76,
                    "word": ".",
                    "probability": 0.9638671875
                }
            ]
        },
        {
            "id": 243,
            "seek": 150350,
            "start": 1501.66,
            "end": 1505.16,
            "text": " So hopefully you get some intuition as to how powerful this attentional representation is and",
            "tokens": [
                50728,
                407,
                4696,
                291,
                483,
                512,
                24002,
                382,
                281,
                577,
                4005,
                341,
                3202,
                304,
                10290,
                307,
                293,
                50904
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22926979725903804,
            "compression_ratio": 1.6765799256505576,
            "no_speech_prob": 0.00037980079650878906,
            "words": [
                {
                    "start": 1501.66,
                    "end": 1501.8,
                    "word": " So",
                    "probability": 0.73681640625
                },
                {
                    "start": 1501.8,
                    "end": 1502.12,
                    "word": " hopefully",
                    "probability": 0.5439453125
                },
                {
                    "start": 1502.12,
                    "end": 1502.24,
                    "word": " you",
                    "probability": 0.7734375
                },
                {
                    "start": 1502.24,
                    "end": 1502.36,
                    "word": " get",
                    "probability": 0.77197265625
                },
                {
                    "start": 1502.36,
                    "end": 1502.54,
                    "word": " some",
                    "probability": 0.8115234375
                },
                {
                    "start": 1502.54,
                    "end": 1502.84,
                    "word": " intuition",
                    "probability": 0.826171875
                },
                {
                    "start": 1502.84,
                    "end": 1503.06,
                    "word": " as",
                    "probability": 0.88427734375
                },
                {
                    "start": 1503.06,
                    "end": 1503.18,
                    "word": " to",
                    "probability": 0.89794921875
                },
                {
                    "start": 1503.18,
                    "end": 1503.32,
                    "word": " how",
                    "probability": 0.892578125
                },
                {
                    "start": 1503.32,
                    "end": 1503.68,
                    "word": " powerful",
                    "probability": 0.84033203125
                },
                {
                    "start": 1503.68,
                    "end": 1503.88,
                    "word": " this",
                    "probability": 0.875
                },
                {
                    "start": 1503.88,
                    "end": 1504.28,
                    "word": " attentional",
                    "probability": 0.8466796875
                },
                {
                    "start": 1504.28,
                    "end": 1504.62,
                    "word": " representation",
                    "probability": 0.76611328125
                },
                {
                    "start": 1504.62,
                    "end": 1505.06,
                    "word": " is",
                    "probability": 0.8896484375
                },
                {
                    "start": 1505.06,
                    "end": 1505.16,
                    "word": " and",
                    "probability": 0.7138671875
                }
            ]
        },
        {
            "id": 244,
            "seek": 150350,
            "start": 1505.16,
            "end": 1511.16,
            "text": " how important it is in machine learning models, getting some what we might call understanding",
            "tokens": [
                50904,
                577,
                1021,
                309,
                307,
                294,
                3479,
                2539,
                5245,
                11,
                1242,
                512,
                437,
                321,
                1062,
                818,
                3701,
                51216
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22926979725903804,
            "compression_ratio": 1.6765799256505576,
            "no_speech_prob": 0.00037980079650878906,
            "words": [
                {
                    "start": 1505.16,
                    "end": 1505.28,
                    "word": " how",
                    "probability": 0.90087890625
                },
                {
                    "start": 1505.28,
                    "end": 1505.64,
                    "word": " important",
                    "probability": 0.76220703125
                },
                {
                    "start": 1505.64,
                    "end": 1506.18,
                    "word": " it",
                    "probability": 0.8994140625
                },
                {
                    "start": 1506.18,
                    "end": 1506.36,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 1506.36,
                    "end": 1506.68,
                    "word": " in",
                    "probability": 0.80322265625
                },
                {
                    "start": 1506.68,
                    "end": 1507.78,
                    "word": " machine",
                    "probability": 0.6396484375
                },
                {
                    "start": 1507.78,
                    "end": 1508.06,
                    "word": " learning",
                    "probability": 0.7861328125
                },
                {
                    "start": 1508.06,
                    "end": 1508.42,
                    "word": " models",
                    "probability": 0.79296875
                },
                {
                    "start": 1508.42,
                    "end": 1508.58,
                    "word": ",",
                    "probability": 0.56982421875
                },
                {
                    "start": 1508.58,
                    "end": 1508.82,
                    "word": " getting",
                    "probability": 0.62060546875
                },
                {
                    "start": 1508.82,
                    "end": 1509.54,
                    "word": " some",
                    "probability": 0.74951171875
                },
                {
                    "start": 1509.54,
                    "end": 1510.22,
                    "word": " what",
                    "probability": 0.258056640625
                },
                {
                    "start": 1510.22,
                    "end": 1510.3,
                    "word": " we",
                    "probability": 0.90087890625
                },
                {
                    "start": 1510.3,
                    "end": 1510.44,
                    "word": " might",
                    "probability": 0.73046875
                },
                {
                    "start": 1510.44,
                    "end": 1510.6,
                    "word": " call",
                    "probability": 0.826171875
                },
                {
                    "start": 1510.6,
                    "end": 1511.16,
                    "word": " understanding",
                    "probability": 0.58251953125
                }
            ]
        },
        {
            "id": 245,
            "seek": 150350,
            "start": 1511.16,
            "end": 1518.41,
            "text": " over the language that they're analyzing. Okay, so I want to put the intuition together",
            "tokens": [
                51216,
                670,
                264,
                2856,
                300,
                436,
                434,
                23663,
                13,
                1033,
                11,
                370,
                286,
                528,
                281,
                829,
                264,
                24002,
                1214,
                51512
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22926979725903804,
            "compression_ratio": 1.6765799256505576,
            "no_speech_prob": 0.00037980079650878906,
            "words": [
                {
                    "start": 1511.16,
                    "end": 1511.52,
                    "word": " over",
                    "probability": 0.7421875
                },
                {
                    "start": 1511.52,
                    "end": 1512.0,
                    "word": " the",
                    "probability": 0.8095703125
                },
                {
                    "start": 1512.0,
                    "end": 1512.22,
                    "word": " language",
                    "probability": 0.57861328125
                },
                {
                    "start": 1512.22,
                    "end": 1512.44,
                    "word": " that",
                    "probability": 0.859375
                },
                {
                    "start": 1512.44,
                    "end": 1512.58,
                    "word": " they're",
                    "probability": 0.8251953125
                },
                {
                    "start": 1512.58,
                    "end": 1512.98,
                    "word": " analyzing",
                    "probability": 0.685546875
                },
                {
                    "start": 1512.98,
                    "end": 1513.52,
                    "word": ".",
                    "probability": 0.98974609375
                },
                {
                    "start": 1516.31,
                    "end": 1516.65,
                    "word": " Okay",
                    "probability": 0.6162109375
                },
                {
                    "start": 1516.65,
                    "end": 1516.85,
                    "word": ",",
                    "probability": 0.77099609375
                },
                {
                    "start": 1516.85,
                    "end": 1517.07,
                    "word": " so",
                    "probability": 0.87548828125
                },
                {
                    "start": 1517.07,
                    "end": 1517.27,
                    "word": " I",
                    "probability": 0.97900390625
                },
                {
                    "start": 1517.27,
                    "end": 1517.33,
                    "word": " want",
                    "probability": 0.49951171875
                },
                {
                    "start": 1517.33,
                    "end": 1517.41,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1517.41,
                    "end": 1517.55,
                    "word": " put",
                    "probability": 0.88427734375
                },
                {
                    "start": 1517.55,
                    "end": 1517.67,
                    "word": " the",
                    "probability": 0.818359375
                },
                {
                    "start": 1517.67,
                    "end": 1517.97,
                    "word": " intuition",
                    "probability": 0.7998046875
                },
                {
                    "start": 1517.97,
                    "end": 1518.41,
                    "word": " together",
                    "probability": 0.77978515625
                }
            ]
        },
        {
            "id": 246,
            "seek": 150350,
            "start": 1518.41,
            "end": 1523.41,
            "text": " around how transformers pertain to the work that YoungHu and I are presenting today.",
            "tokens": [
                51512,
                926,
                577,
                4088,
                433,
                13269,
                491,
                281,
                264,
                589,
                300,
                8160,
                39,
                84,
                293,
                286,
                366,
                15578,
                965,
                13,
                51728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22926979725903804,
            "compression_ratio": 1.6765799256505576,
            "no_speech_prob": 0.00037980079650878906,
            "words": [
                {
                    "start": 1518.41,
                    "end": 1518.89,
                    "word": " around",
                    "probability": 0.85302734375
                },
                {
                    "start": 1518.89,
                    "end": 1519.11,
                    "word": " how",
                    "probability": 0.89453125
                },
                {
                    "start": 1519.11,
                    "end": 1519.79,
                    "word": " transformers",
                    "probability": 0.80712890625
                },
                {
                    "start": 1519.79,
                    "end": 1520.41,
                    "word": " pertain",
                    "probability": 0.905029296875
                },
                {
                    "start": 1520.41,
                    "end": 1520.61,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 1520.61,
                    "end": 1520.91,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 1520.91,
                    "end": 1521.31,
                    "word": " work",
                    "probability": 0.7783203125
                },
                {
                    "start": 1521.31,
                    "end": 1521.49,
                    "word": " that",
                    "probability": 0.88623046875
                },
                {
                    "start": 1521.49,
                    "end": 1521.77,
                    "word": " YoungHu",
                    "probability": 0.4559733072916667
                },
                {
                    "start": 1521.77,
                    "end": 1521.91,
                    "word": " and",
                    "probability": 0.89453125
                },
                {
                    "start": 1521.91,
                    "end": 1522.05,
                    "word": " I",
                    "probability": 0.9990234375
                },
                {
                    "start": 1522.05,
                    "end": 1522.23,
                    "word": " are",
                    "probability": 0.89404296875
                },
                {
                    "start": 1522.23,
                    "end": 1522.51,
                    "word": " presenting",
                    "probability": 0.732421875
                },
                {
                    "start": 1522.51,
                    "end": 1522.87,
                    "word": " today",
                    "probability": 0.79345703125
                },
                {
                    "start": 1522.87,
                    "end": 1523.41,
                    "word": ".",
                    "probability": 0.9892578125
                }
            ]
        },
        {
            "id": 247,
            "seek": 152644,
            "start": 1524.23,
            "end": 1531.35,
            "text": " So basically, what we're going to do in our phishing model is embed an email as a sequence",
            "tokens": [
                50412,
                407,
                1936,
                11,
                437,
                321,
                434,
                516,
                281,
                360,
                294,
                527,
                903,
                3807,
                2316,
                307,
                12240,
                364,
                3796,
                382,
                257,
                8310,
                50752
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2566287982644457,
            "compression_ratio": 1.5555555555555556,
            "no_speech_prob": 8.088350296020508e-05,
            "words": [
                {
                    "start": 1524.23,
                    "end": 1524.89,
                    "word": " So",
                    "probability": 0.66943359375
                },
                {
                    "start": 1524.89,
                    "end": 1525.25,
                    "word": " basically",
                    "probability": 0.7578125
                },
                {
                    "start": 1525.25,
                    "end": 1525.87,
                    "word": ",",
                    "probability": 0.5966796875
                },
                {
                    "start": 1525.87,
                    "end": 1526.55,
                    "word": " what",
                    "probability": 0.880859375
                },
                {
                    "start": 1526.55,
                    "end": 1526.85,
                    "word": " we're",
                    "probability": 0.91357421875
                },
                {
                    "start": 1526.85,
                    "end": 1526.99,
                    "word": " going",
                    "probability": 0.2978515625
                },
                {
                    "start": 1526.99,
                    "end": 1527.05,
                    "word": " to",
                    "probability": 0.9052734375
                },
                {
                    "start": 1527.05,
                    "end": 1527.19,
                    "word": " do",
                    "probability": 0.91552734375
                },
                {
                    "start": 1527.19,
                    "end": 1527.31,
                    "word": " in",
                    "probability": 0.80029296875
                },
                {
                    "start": 1527.31,
                    "end": 1527.43,
                    "word": " our",
                    "probability": 0.77685546875
                },
                {
                    "start": 1527.43,
                    "end": 1527.83,
                    "word": " phishing",
                    "probability": 0.802001953125
                },
                {
                    "start": 1527.83,
                    "end": 1528.13,
                    "word": " model",
                    "probability": 0.8203125
                },
                {
                    "start": 1528.13,
                    "end": 1528.69,
                    "word": " is",
                    "probability": 0.8056640625
                },
                {
                    "start": 1528.69,
                    "end": 1529.31,
                    "word": " embed",
                    "probability": 0.83251953125
                },
                {
                    "start": 1529.31,
                    "end": 1529.89,
                    "word": " an",
                    "probability": 0.7802734375
                },
                {
                    "start": 1529.89,
                    "end": 1530.23,
                    "word": " email",
                    "probability": 0.87548828125
                },
                {
                    "start": 1530.23,
                    "end": 1530.85,
                    "word": " as",
                    "probability": 0.84521484375
                },
                {
                    "start": 1530.97,
                    "end": 1531.35,
                    "word": " a sequence",
                    "probability": 0.9072265625
                }
            ]
        },
        {
            "id": 248,
            "seek": 152644,
            "start": 1531.35,
            "end": 1539.37,
            "text": " of embedded character sequence vectors. And then we're going to run that through a series of",
            "tokens": [
                50752,
                295,
                16741,
                2517,
                8310,
                18875,
                13,
                400,
                550,
                321,
                434,
                516,
                281,
                1190,
                300,
                807,
                257,
                2638,
                295,
                51148
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2566287982644457,
            "compression_ratio": 1.5555555555555556,
            "no_speech_prob": 8.088350296020508e-05,
            "words": [
                {
                    "start": 1531.35,
                    "end": 1531.79,
                    "word": " of",
                    "probability": 0.8955078125
                },
                {
                    "start": 1531.79,
                    "end": 1533.11,
                    "word": " embedded",
                    "probability": 0.57470703125
                },
                {
                    "start": 1533.11,
                    "end": 1535.45,
                    "word": " character",
                    "probability": 0.65380859375
                },
                {
                    "start": 1535.45,
                    "end": 1535.83,
                    "word": " sequence",
                    "probability": 0.8974609375
                },
                {
                    "start": 1535.83,
                    "end": 1536.23,
                    "word": " vectors",
                    "probability": 0.89599609375
                },
                {
                    "start": 1536.23,
                    "end": 1536.65,
                    "word": ".",
                    "probability": 0.56005859375
                },
                {
                    "start": 1536.65,
                    "end": 1537.23,
                    "word": " And",
                    "probability": 0.8447265625
                },
                {
                    "start": 1537.23,
                    "end": 1537.43,
                    "word": " then",
                    "probability": 0.72021484375
                },
                {
                    "start": 1537.43,
                    "end": 1537.83,
                    "word": " we're",
                    "probability": 0.875
                },
                {
                    "start": 1537.83,
                    "end": 1537.89,
                    "word": " going",
                    "probability": 0.79931640625
                },
                {
                    "start": 1537.89,
                    "end": 1537.99,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 1537.99,
                    "end": 1538.21,
                    "word": " run",
                    "probability": 0.900390625
                },
                {
                    "start": 1538.21,
                    "end": 1538.43,
                    "word": " that",
                    "probability": 0.8828125
                },
                {
                    "start": 1538.43,
                    "end": 1538.63,
                    "word": " through",
                    "probability": 0.79345703125
                },
                {
                    "start": 1538.83,
                    "end": 1539.15,
                    "word": " a series",
                    "probability": 0.82080078125
                },
                {
                    "start": 1539.15,
                    "end": 1539.37,
                    "word": " of",
                    "probability": 0.90478515625
                }
            ]
        },
        {
            "id": 249,
            "seek": 152644,
            "start": 1539.37,
            "end": 1546.35,
            "text": " transformer blocks, like what I just showed, that are going to create a very refined attentional",
            "tokens": [
                51148,
                31782,
                8474,
                11,
                411,
                437,
                286,
                445,
                4712,
                11,
                300,
                366,
                516,
                281,
                1884,
                257,
                588,
                26201,
                3202,
                304,
                51500
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2566287982644457,
            "compression_ratio": 1.5555555555555556,
            "no_speech_prob": 8.088350296020508e-05,
            "words": [
                {
                    "start": 1539.37,
                    "end": 1539.79,
                    "word": " transformer",
                    "probability": 0.794921875
                },
                {
                    "start": 1539.79,
                    "end": 1540.19,
                    "word": " blocks",
                    "probability": 0.81201171875
                },
                {
                    "start": 1540.19,
                    "end": 1540.23,
                    "word": ",",
                    "probability": 0.464599609375
                },
                {
                    "start": 1540.23,
                    "end": 1540.37,
                    "word": " like",
                    "probability": 0.87646484375
                },
                {
                    "start": 1540.37,
                    "end": 1540.49,
                    "word": " what",
                    "probability": 0.87890625
                },
                {
                    "start": 1540.49,
                    "end": 1540.59,
                    "word": " I",
                    "probability": 0.99755859375
                },
                {
                    "start": 1540.59,
                    "end": 1540.73,
                    "word": " just",
                    "probability": 0.8916015625
                },
                {
                    "start": 1540.73,
                    "end": 1541.03,
                    "word": " showed",
                    "probability": 0.783203125
                },
                {
                    "start": 1541.03,
                    "end": 1541.39,
                    "word": ",",
                    "probability": 0.78857421875
                },
                {
                    "start": 1541.39,
                    "end": 1541.89,
                    "word": " that",
                    "probability": 0.88037109375
                },
                {
                    "start": 1541.89,
                    "end": 1542.05,
                    "word": " are",
                    "probability": 0.89501953125
                },
                {
                    "start": 1542.05,
                    "end": 1542.17,
                    "word": " going",
                    "probability": 0.83837890625
                },
                {
                    "start": 1542.17,
                    "end": 1542.93,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 1542.93,
                    "end": 1543.71,
                    "word": " create",
                    "probability": 0.83154296875
                },
                {
                    "start": 1543.89,
                    "end": 1544.09,
                    "word": " a very",
                    "probability": 0.74853515625
                },
                {
                    "start": 1544.09,
                    "end": 1544.93,
                    "word": " refined",
                    "probability": 0.85888671875
                },
                {
                    "start": 1544.93,
                    "end": 1546.35,
                    "word": " attentional",
                    "probability": 0.79052734375
                }
            ]
        },
        {
            "id": 250,
            "seek": 155028,
            "start": 1546.35,
            "end": 1554.79,
            "text": " representation of the word sequences, and then produce these contextual embeddings that get at",
            "tokens": [
                50364,
                10290,
                295,
                264,
                1349,
                22978,
                11,
                293,
                550,
                5258,
                613,
                35526,
                12240,
                29432,
                300,
                483,
                412,
                50784
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2122093044048132,
            "compression_ratio": 1.6905829596412556,
            "no_speech_prob": 0.005817413330078125,
            "words": [
                {
                    "start": 1546.35,
                    "end": 1546.75,
                    "word": " representation",
                    "probability": 0.65869140625
                },
                {
                    "start": 1546.75,
                    "end": 1548.97,
                    "word": " of",
                    "probability": 0.8466796875
                },
                {
                    "start": 1548.97,
                    "end": 1550.13,
                    "word": " the",
                    "probability": 0.7998046875
                },
                {
                    "start": 1550.13,
                    "end": 1550.35,
                    "word": " word",
                    "probability": 0.7802734375
                },
                {
                    "start": 1550.35,
                    "end": 1550.81,
                    "word": " sequences",
                    "probability": 0.90380859375
                },
                {
                    "start": 1550.81,
                    "end": 1551.43,
                    "word": ",",
                    "probability": 0.255615234375
                },
                {
                    "start": 1551.43,
                    "end": 1551.91,
                    "word": " and",
                    "probability": 0.8515625
                },
                {
                    "start": 1551.91,
                    "end": 1552.03,
                    "word": " then",
                    "probability": 0.7265625
                },
                {
                    "start": 1552.03,
                    "end": 1552.31,
                    "word": " produce",
                    "probability": 0.75
                },
                {
                    "start": 1552.31,
                    "end": 1552.49,
                    "word": " these",
                    "probability": 0.685546875
                },
                {
                    "start": 1552.49,
                    "end": 1552.93,
                    "word": " contextual",
                    "probability": 0.86279296875
                },
                {
                    "start": 1552.93,
                    "end": 1553.43,
                    "word": " embeddings",
                    "probability": 0.853515625
                },
                {
                    "start": 1553.43,
                    "end": 1553.73,
                    "word": " that",
                    "probability": 0.85791015625
                },
                {
                    "start": 1553.73,
                    "end": 1554.67,
                    "word": " get",
                    "probability": 0.724609375
                },
                {
                    "start": 1554.67,
                    "end": 1554.79,
                    "word": " at",
                    "probability": 0.82470703125
                }
            ]
        },
        {
            "id": 251,
            "seek": 155028,
            "start": 1554.79,
            "end": 1560.49,
            "text": " the meaning of the words in the context in which they appear. And then finally, our network is",
            "tokens": [
                50784,
                264,
                3620,
                295,
                264,
                2283,
                294,
                264,
                4319,
                294,
                597,
                436,
                4204,
                13,
                400,
                550,
                2721,
                11,
                527,
                3209,
                307,
                51068
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2122093044048132,
            "compression_ratio": 1.6905829596412556,
            "no_speech_prob": 0.005817413330078125,
            "words": [
                {
                    "start": 1554.79,
                    "end": 1554.91,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 1554.91,
                    "end": 1555.17,
                    "word": " meaning",
                    "probability": 0.77978515625
                },
                {
                    "start": 1555.17,
                    "end": 1555.49,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 1555.49,
                    "end": 1555.61,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 1555.61,
                    "end": 1555.85,
                    "word": " words",
                    "probability": 0.72607421875
                },
                {
                    "start": 1555.85,
                    "end": 1555.99,
                    "word": " in",
                    "probability": 0.386474609375
                },
                {
                    "start": 1555.99,
                    "end": 1556.11,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 1556.11,
                    "end": 1556.45,
                    "word": " context",
                    "probability": 0.74462890625
                },
                {
                    "start": 1556.45,
                    "end": 1556.59,
                    "word": " in",
                    "probability": 0.880859375
                },
                {
                    "start": 1556.59,
                    "end": 1556.71,
                    "word": " which",
                    "probability": 0.91064453125
                },
                {
                    "start": 1556.71,
                    "end": 1556.85,
                    "word": " they",
                    "probability": 0.736328125
                },
                {
                    "start": 1556.85,
                    "end": 1557.21,
                    "word": " appear",
                    "probability": 0.69580078125
                },
                {
                    "start": 1557.21,
                    "end": 1557.65,
                    "word": ".",
                    "probability": 0.8662109375
                },
                {
                    "start": 1557.87,
                    "end": 1557.93,
                    "word": " And",
                    "probability": 0.78515625
                },
                {
                    "start": 1557.93,
                    "end": 1558.07,
                    "word": " then",
                    "probability": 0.71044921875
                },
                {
                    "start": 1558.07,
                    "end": 1558.41,
                    "word": " finally",
                    "probability": 0.73291015625
                },
                {
                    "start": 1558.41,
                    "end": 1559.03,
                    "word": ",",
                    "probability": 0.8056640625
                },
                {
                    "start": 1559.03,
                    "end": 1559.93,
                    "word": " our",
                    "probability": 0.8134765625
                },
                {
                    "start": 1559.93,
                    "end": 1560.29,
                    "word": " network",
                    "probability": 0.82568359375
                },
                {
                    "start": 1560.29,
                    "end": 1560.49,
                    "word": " is",
                    "probability": 0.9091796875
                }
            ]
        },
        {
            "id": 252,
            "seek": 155028,
            "start": 1560.49,
            "end": 1565.79,
            "text": " going to solve a classification task, say whether or not the email is a phishing email or not.",
            "tokens": [
                51068,
                516,
                281,
                5039,
                257,
                21538,
                5633,
                11,
                584,
                1968,
                420,
                406,
                264,
                3796,
                307,
                257,
                903,
                3807,
                3796,
                420,
                406,
                13,
                51308
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2122093044048132,
            "compression_ratio": 1.6905829596412556,
            "no_speech_prob": 0.005817413330078125,
            "words": [
                {
                    "start": 1560.49,
                    "end": 1560.71,
                    "word": " going",
                    "probability": 0.85986328125
                },
                {
                    "start": 1560.71,
                    "end": 1561.23,
                    "word": " to",
                    "probability": 0.9013671875
                },
                {
                    "start": 1561.23,
                    "end": 1561.65,
                    "word": " solve",
                    "probability": 0.830078125
                },
                {
                    "start": 1561.79,
                    "end": 1562.17,
                    "word": " a classification",
                    "probability": 0.92724609375
                },
                {
                    "start": 1562.17,
                    "end": 1562.81,
                    "word": " task",
                    "probability": 0.79736328125
                },
                {
                    "start": 1562.81,
                    "end": 1563.03,
                    "word": ",",
                    "probability": 0.70166015625
                },
                {
                    "start": 1563.03,
                    "end": 1563.23,
                    "word": " say",
                    "probability": 0.7724609375
                },
                {
                    "start": 1563.23,
                    "end": 1563.39,
                    "word": " whether",
                    "probability": 0.7314453125
                },
                {
                    "start": 1563.39,
                    "end": 1563.53,
                    "word": " or",
                    "probability": 0.8818359375
                },
                {
                    "start": 1563.53,
                    "end": 1563.63,
                    "word": " not",
                    "probability": 0.88916015625
                },
                {
                    "start": 1563.63,
                    "end": 1563.77,
                    "word": " the",
                    "probability": 0.7890625
                },
                {
                    "start": 1563.77,
                    "end": 1563.95,
                    "word": " email",
                    "probability": 0.875
                },
                {
                    "start": 1563.95,
                    "end": 1564.11,
                    "word": " is",
                    "probability": 0.88818359375
                },
                {
                    "start": 1564.23,
                    "end": 1564.41,
                    "word": " a phishing",
                    "probability": 0.89111328125
                },
                {
                    "start": 1564.41,
                    "end": 1564.69,
                    "word": " email",
                    "probability": 0.89794921875
                },
                {
                    "start": 1564.69,
                    "end": 1565.05,
                    "word": " or",
                    "probability": 0.8408203125
                },
                {
                    "start": 1565.05,
                    "end": 1565.35,
                    "word": " not",
                    "probability": 0.89453125
                },
                {
                    "start": 1565.35,
                    "end": 1565.79,
                    "word": ".",
                    "probability": 0.94873046875
                }
            ]
        },
        {
            "id": 253,
            "seek": 155028,
            "start": 1566.31,
            "end": 1570.19,
            "text": " How all that magic works, how the network gets trained, which there's some tricks there that",
            "tokens": [
                51360,
                1012,
                439,
                300,
                5585,
                1985,
                11,
                577,
                264,
                3209,
                2170,
                8895,
                11,
                597,
                456,
                311,
                512,
                11733,
                456,
                300,
                51552
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2122093044048132,
            "compression_ratio": 1.6905829596412556,
            "no_speech_prob": 0.005817413330078125,
            "words": [
                {
                    "start": 1566.31,
                    "end": 1566.51,
                    "word": " How",
                    "probability": 0.8857421875
                },
                {
                    "start": 1566.51,
                    "end": 1566.63,
                    "word": " all",
                    "probability": 0.82470703125
                },
                {
                    "start": 1566.63,
                    "end": 1566.75,
                    "word": " that",
                    "probability": 0.85888671875
                },
                {
                    "start": 1566.75,
                    "end": 1566.99,
                    "word": " magic",
                    "probability": 0.8154296875
                },
                {
                    "start": 1566.99,
                    "end": 1567.45,
                    "word": " works",
                    "probability": 0.73291015625
                },
                {
                    "start": 1567.45,
                    "end": 1567.67,
                    "word": ",",
                    "probability": 0.908203125
                },
                {
                    "start": 1567.67,
                    "end": 1567.79,
                    "word": " how",
                    "probability": 0.8916015625
                },
                {
                    "start": 1567.79,
                    "end": 1567.87,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 1567.87,
                    "end": 1568.05,
                    "word": " network",
                    "probability": 0.82177734375
                },
                {
                    "start": 1568.05,
                    "end": 1568.27,
                    "word": " gets",
                    "probability": 0.79736328125
                },
                {
                    "start": 1568.27,
                    "end": 1568.65,
                    "word": " trained",
                    "probability": 0.81787109375
                },
                {
                    "start": 1568.65,
                    "end": 1568.99,
                    "word": ",",
                    "probability": 0.849609375
                },
                {
                    "start": 1568.99,
                    "end": 1569.21,
                    "word": " which",
                    "probability": 0.83056640625
                },
                {
                    "start": 1569.21,
                    "end": 1569.49,
                    "word": " there's",
                    "probability": 0.824951171875
                },
                {
                    "start": 1569.49,
                    "end": 1569.65,
                    "word": " some",
                    "probability": 0.8134765625
                },
                {
                    "start": 1569.65,
                    "end": 1569.87,
                    "word": " tricks",
                    "probability": 0.705078125
                },
                {
                    "start": 1569.87,
                    "end": 1570.07,
                    "word": " there",
                    "probability": 0.853515625
                },
                {
                    "start": 1570.07,
                    "end": 1570.19,
                    "word": " that",
                    "probability": 0.8798828125
                }
            ]
        },
        {
            "id": 254,
            "seek": 157094,
            "start": 1570.19,
            "end": 1576.17,
            "text": " are really cool. I'll leave to YoungHu, and hope a lot of this makes sense, and happy to",
            "tokens": [
                50364,
                366,
                534,
                1627,
                13,
                286,
                603,
                1856,
                281,
                8160,
                39,
                84,
                11,
                293,
                1454,
                257,
                688,
                295,
                341,
                1669,
                2020,
                11,
                293,
                2055,
                281,
                50660
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25973732056825055,
            "compression_ratio": 1.5174418604651163,
            "no_speech_prob": 0.053955078125,
            "words": [
                {
                    "start": 1570.19,
                    "end": 1570.25,
                    "word": " are",
                    "probability": 0.086669921875
                },
                {
                    "start": 1570.25,
                    "end": 1570.41,
                    "word": " really",
                    "probability": 0.77783203125
                },
                {
                    "start": 1570.41,
                    "end": 1570.75,
                    "word": " cool",
                    "probability": 0.884765625
                },
                {
                    "start": 1570.75,
                    "end": 1571.31,
                    "word": ".",
                    "probability": 0.70458984375
                },
                {
                    "start": 1571.77,
                    "end": 1572.33,
                    "word": " I'll",
                    "probability": 0.954833984375
                },
                {
                    "start": 1572.33,
                    "end": 1572.47,
                    "word": " leave",
                    "probability": 0.9072265625
                },
                {
                    "start": 1572.47,
                    "end": 1572.61,
                    "word": " to",
                    "probability": 0.724609375
                },
                {
                    "start": 1572.61,
                    "end": 1572.91,
                    "word": " YoungHu",
                    "probability": 0.238037109375
                },
                {
                    "start": 1572.91,
                    "end": 1573.43,
                    "word": ",",
                    "probability": 0.2191162109375
                },
                {
                    "start": 1573.43,
                    "end": 1573.69,
                    "word": " and",
                    "probability": 0.86572265625
                },
                {
                    "start": 1573.69,
                    "end": 1574.63,
                    "word": " hope",
                    "probability": 0.64697265625
                },
                {
                    "start": 1574.73,
                    "end": 1574.79,
                    "word": " a lot",
                    "probability": 0.90576171875
                },
                {
                    "start": 1574.79,
                    "end": 1574.87,
                    "word": " of",
                    "probability": 0.8994140625
                },
                {
                    "start": 1574.87,
                    "end": 1574.97,
                    "word": " this",
                    "probability": 0.87548828125
                },
                {
                    "start": 1574.97,
                    "end": 1575.17,
                    "word": " makes",
                    "probability": 0.6767578125
                },
                {
                    "start": 1575.17,
                    "end": 1575.49,
                    "word": " sense",
                    "probability": 0.7470703125
                },
                {
                    "start": 1575.49,
                    "end": 1575.69,
                    "word": ",",
                    "probability": 0.335693359375
                },
                {
                    "start": 1575.69,
                    "end": 1575.81,
                    "word": " and",
                    "probability": 0.88818359375
                },
                {
                    "start": 1575.81,
                    "end": 1576.03,
                    "word": " happy",
                    "probability": 0.74755859375
                },
                {
                    "start": 1576.03,
                    "end": 1576.17,
                    "word": " to",
                    "probability": 0.904296875
                }
            ]
        },
        {
            "id": 255,
            "seek": 157094,
            "start": 1576.17,
            "end": 1580.55,
            "text": " take questions about my piece of this presentation later at the end of the talk.",
            "tokens": [
                50660,
                747,
                1651,
                466,
                452,
                2522,
                295,
                341,
                5860,
                1780,
                412,
                264,
                917,
                295,
                264,
                751,
                13,
                50852
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25973732056825055,
            "compression_ratio": 1.5174418604651163,
            "no_speech_prob": 0.053955078125,
            "words": [
                {
                    "start": 1576.17,
                    "end": 1576.33,
                    "word": " take",
                    "probability": 0.8056640625
                },
                {
                    "start": 1576.33,
                    "end": 1576.63,
                    "word": " questions",
                    "probability": 0.91796875
                },
                {
                    "start": 1576.63,
                    "end": 1576.85,
                    "word": " about",
                    "probability": 0.8125
                },
                {
                    "start": 1576.85,
                    "end": 1577.03,
                    "word": " my",
                    "probability": 0.91015625
                },
                {
                    "start": 1577.03,
                    "end": 1577.23,
                    "word": " piece",
                    "probability": 0.7578125
                },
                {
                    "start": 1577.23,
                    "end": 1577.33,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1577.33,
                    "end": 1577.49,
                    "word": " this",
                    "probability": 0.8759765625
                },
                {
                    "start": 1577.49,
                    "end": 1577.89,
                    "word": " presentation",
                    "probability": 0.77099609375
                },
                {
                    "start": 1577.89,
                    "end": 1579.13,
                    "word": " later",
                    "probability": 0.82421875
                },
                {
                    "start": 1579.13,
                    "end": 1579.37,
                    "word": " at",
                    "probability": 0.7958984375
                },
                {
                    "start": 1579.37,
                    "end": 1579.45,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 1579.45,
                    "end": 1579.53,
                    "word": " end",
                    "probability": 0.79931640625
                },
                {
                    "start": 1579.53,
                    "end": 1579.61,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 1579.61,
                    "end": 1579.71,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 1579.71,
                    "end": 1579.99,
                    "word": " talk",
                    "probability": 0.81982421875
                },
                {
                    "start": 1579.99,
                    "end": 1580.55,
                    "word": ".",
                    "probability": 0.990234375
                }
            ]
        },
        {
            "id": 256,
            "seek": 157094,
            "start": 1581.97,
            "end": 1590.85,
            "text": " Thank you, Joshi. Let me continue the second part of our talk. The second part will include",
            "tokens": [
                50964,
                1044,
                291,
                11,
                9785,
                72,
                13,
                961,
                385,
                2354,
                264,
                1150,
                644,
                295,
                527,
                751,
                13,
                440,
                1150,
                644,
                486,
                4090,
                51392
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25973732056825055,
            "compression_ratio": 1.5174418604651163,
            "no_speech_prob": 0.053955078125,
            "words": [
                {
                    "start": 1581.97,
                    "end": 1582.53,
                    "word": " Thank",
                    "probability": 0.492919921875
                },
                {
                    "start": 1582.53,
                    "end": 1582.83,
                    "word": " you",
                    "probability": 0.8935546875
                },
                {
                    "start": 1582.83,
                    "end": 1583.21,
                    "word": ",",
                    "probability": 0.8564453125
                },
                {
                    "start": 1583.21,
                    "end": 1583.63,
                    "word": " Joshi",
                    "probability": 0.5263671875
                },
                {
                    "start": 1583.63,
                    "end": 1584.15,
                    "word": ".",
                    "probability": 0.958984375
                },
                {
                    "start": 1584.15,
                    "end": 1584.61,
                    "word": " Let",
                    "probability": 0.87451171875
                },
                {
                    "start": 1584.61,
                    "end": 1584.77,
                    "word": " me",
                    "probability": 0.9033203125
                },
                {
                    "start": 1584.77,
                    "end": 1585.31,
                    "word": " continue",
                    "probability": 0.822265625
                },
                {
                    "start": 1585.31,
                    "end": 1585.65,
                    "word": " the",
                    "probability": 0.53759765625
                },
                {
                    "start": 1585.65,
                    "end": 1585.97,
                    "word": " second",
                    "probability": 0.7783203125
                },
                {
                    "start": 1585.97,
                    "end": 1586.35,
                    "word": " part",
                    "probability": 0.828125
                },
                {
                    "start": 1586.35,
                    "end": 1586.61,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 1586.61,
                    "end": 1587.07,
                    "word": " our",
                    "probability": 0.81982421875
                },
                {
                    "start": 1587.07,
                    "end": 1587.73,
                    "word": " talk",
                    "probability": 0.8017578125
                },
                {
                    "start": 1587.73,
                    "end": 1588.29,
                    "word": ".",
                    "probability": 0.98681640625
                },
                {
                    "start": 1588.65,
                    "end": 1588.91,
                    "word": " The",
                    "probability": 0.82080078125
                },
                {
                    "start": 1588.91,
                    "end": 1589.19,
                    "word": " second",
                    "probability": 0.7958984375
                },
                {
                    "start": 1589.19,
                    "end": 1589.61,
                    "word": " part",
                    "probability": 0.8349609375
                },
                {
                    "start": 1589.61,
                    "end": 1590.11,
                    "word": " will",
                    "probability": 0.8134765625
                },
                {
                    "start": 1590.11,
                    "end": 1590.85,
                    "word": " include",
                    "probability": 0.75634765625
                }
            ]
        },
        {
            "id": 257,
            "seek": 159934,
            "start": 1591.57,
            "end": 1599.25,
            "text": " our design decisions for CatBot and performance result.",
            "tokens": [
                50412,
                527,
                1715,
                5327,
                337,
                9565,
                33,
                310,
                293,
                3389,
                1874,
                13,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24007937075599792,
            "compression_ratio": 1.4024390243902438,
            "no_speech_prob": 0.4287109375,
            "words": [
                {
                    "start": 1591.57,
                    "end": 1592.09,
                    "word": " our",
                    "probability": 0.1593017578125
                },
                {
                    "start": 1592.09,
                    "end": 1592.69,
                    "word": " design",
                    "probability": 0.861328125
                },
                {
                    "start": 1592.69,
                    "end": 1593.23,
                    "word": " decisions",
                    "probability": 0.75830078125
                },
                {
                    "start": 1593.23,
                    "end": 1594.01,
                    "word": " for",
                    "probability": 0.8798828125
                },
                {
                    "start": 1594.01,
                    "end": 1594.75,
                    "word": " CatBot",
                    "probability": 0.6673177083333334
                },
                {
                    "start": 1594.75,
                    "end": 1596.11,
                    "word": " and",
                    "probability": 0.728515625
                },
                {
                    "start": 1596.11,
                    "end": 1597.59,
                    "word": " performance",
                    "probability": 0.65283203125
                },
                {
                    "start": 1597.59,
                    "end": 1598.21,
                    "word": " result",
                    "probability": 0.320556640625
                },
                {
                    "start": 1598.21,
                    "end": 1599.25,
                    "word": ".",
                    "probability": 0.73095703125
                }
            ]
        },
        {
            "id": 258,
            "seek": 159934,
            "start": 1599.83,
            "end": 1608.13,
            "text": " CatBot is the name of our email model, context-aware tiny bot. The model size is tiny,",
            "tokens": [
                50812,
                9565,
                33,
                310,
                307,
                264,
                1315,
                295,
                527,
                3796,
                2316,
                11,
                4319,
                12,
                17074,
                5870,
                10592,
                13,
                440,
                2316,
                2744,
                307,
                5870,
                11,
                51220
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24007937075599792,
            "compression_ratio": 1.4024390243902438,
            "no_speech_prob": 0.4287109375,
            "words": [
                {
                    "start": 1599.83,
                    "end": 1600.67,
                    "word": " CatBot",
                    "probability": 0.94091796875
                },
                {
                    "start": 1600.67,
                    "end": 1601.25,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 1601.25,
                    "end": 1601.49,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1601.49,
                    "end": 1601.77,
                    "word": " name",
                    "probability": 0.82373046875
                },
                {
                    "start": 1601.77,
                    "end": 1602.11,
                    "word": " of",
                    "probability": 0.904296875
                },
                {
                    "start": 1602.11,
                    "end": 1602.45,
                    "word": " our",
                    "probability": 0.8154296875
                },
                {
                    "start": 1602.45,
                    "end": 1602.79,
                    "word": " email",
                    "probability": 0.354736328125
                },
                {
                    "start": 1602.79,
                    "end": 1603.19,
                    "word": " model",
                    "probability": 0.83349609375
                },
                {
                    "start": 1603.19,
                    "end": 1603.67,
                    "word": ",",
                    "probability": 0.49951171875
                },
                {
                    "start": 1603.67,
                    "end": 1604.17,
                    "word": " context",
                    "probability": 0.7216796875
                },
                {
                    "start": 1604.17,
                    "end": 1604.75,
                    "word": "-aware",
                    "probability": 0.76123046875
                },
                {
                    "start": 1604.75,
                    "end": 1605.39,
                    "word": " tiny",
                    "probability": 0.4921875
                },
                {
                    "start": 1605.39,
                    "end": 1605.69,
                    "word": " bot",
                    "probability": 0.56494140625
                },
                {
                    "start": 1605.69,
                    "end": 1606.31,
                    "word": ".",
                    "probability": 0.8896484375
                },
                {
                    "start": 1606.31,
                    "end": 1606.53,
                    "word": " The",
                    "probability": 0.775390625
                },
                {
                    "start": 1606.53,
                    "end": 1606.79,
                    "word": " model",
                    "probability": 0.8046875
                },
                {
                    "start": 1606.79,
                    "end": 1607.15,
                    "word": " size",
                    "probability": 0.78515625
                },
                {
                    "start": 1607.15,
                    "end": 1607.35,
                    "word": " is",
                    "probability": 0.91015625
                },
                {
                    "start": 1607.35,
                    "end": 1607.79,
                    "word": " tiny",
                    "probability": 0.73974609375
                },
                {
                    "start": 1607.79,
                    "end": 1608.13,
                    "word": ",",
                    "probability": 0.69287109375
                }
            ]
        },
        {
            "id": 259,
            "seek": 159934,
            "start": 1608.13,
            "end": 1621.58,
            "text": " but it is mighty bot. Modern NLP models all have a nice and friendly name. For example,",
            "tokens": [
                51220,
                457,
                309,
                307,
                21556,
                10592,
                13,
                19814,
                426,
                45196,
                5245,
                439,
                362,
                257,
                1481,
                293,
                9208,
                1315,
                13,
                1171,
                1365,
                11,
                51760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24007937075599792,
            "compression_ratio": 1.4024390243902438,
            "no_speech_prob": 0.4287109375,
            "words": [
                {
                    "start": 1608.13,
                    "end": 1608.37,
                    "word": " but",
                    "probability": 0.8974609375
                },
                {
                    "start": 1608.37,
                    "end": 1608.61,
                    "word": " it",
                    "probability": 0.8935546875
                },
                {
                    "start": 1608.61,
                    "end": 1608.77,
                    "word": " is",
                    "probability": 0.8837890625
                },
                {
                    "start": 1608.77,
                    "end": 1609.27,
                    "word": " mighty",
                    "probability": 0.5576171875
                },
                {
                    "start": 1609.27,
                    "end": 1609.91,
                    "word": " bot",
                    "probability": 0.80126953125
                },
                {
                    "start": 1609.91,
                    "end": 1610.95,
                    "word": ".",
                    "probability": 0.95166015625
                },
                {
                    "start": 1614.04,
                    "end": 1614.56,
                    "word": " Modern",
                    "probability": 0.79541015625
                },
                {
                    "start": 1614.56,
                    "end": 1615.26,
                    "word": " NLP",
                    "probability": 0.97607421875
                },
                {
                    "start": 1615.26,
                    "end": 1615.84,
                    "word": " models",
                    "probability": 0.77099609375
                },
                {
                    "start": 1615.84,
                    "end": 1616.36,
                    "word": " all",
                    "probability": 0.88623046875
                },
                {
                    "start": 1616.36,
                    "end": 1616.88,
                    "word": " have",
                    "probability": 0.884765625
                },
                {
                    "start": 1617.24,
                    "end": 1617.62,
                    "word": " a nice",
                    "probability": 0.7861328125
                },
                {
                    "start": 1617.62,
                    "end": 1617.96,
                    "word": " and",
                    "probability": 0.87255859375
                },
                {
                    "start": 1617.96,
                    "end": 1618.58,
                    "word": " friendly",
                    "probability": 0.62060546875
                },
                {
                    "start": 1618.58,
                    "end": 1619.42,
                    "word": " name",
                    "probability": 0.802734375
                },
                {
                    "start": 1619.42,
                    "end": 1620.28,
                    "word": ".",
                    "probability": 0.9443359375
                },
                {
                    "start": 1620.28,
                    "end": 1620.64,
                    "word": " For",
                    "probability": 0.89013671875
                },
                {
                    "start": 1620.64,
                    "end": 1621.06,
                    "word": " example",
                    "probability": 0.93017578125
                },
                {
                    "start": 1621.06,
                    "end": 1621.58,
                    "word": ",",
                    "probability": 0.96923828125
                }
            ]
        },
        {
            "id": 260,
            "seek": 162574,
            "start": 1622.15,
            "end": 1631.12,
            "text": " Elmer was introduced in 2018. The model used bidirectional LSTMs to",
            "tokens": [
                50400,
                2699,
                936,
                390,
                7268,
                294,
                6096,
                13,
                440,
                2316,
                1143,
                12957,
                621,
                41048,
                441,
                6840,
                26386,
                281,
                50832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3098958263210222,
            "compression_ratio": 1.3006535947712419,
            "no_speech_prob": 0.0023441314697265625,
            "words": [
                {
                    "start": 1622.15,
                    "end": 1622.74,
                    "word": " Elmer",
                    "probability": 0.3756103515625
                },
                {
                    "start": 1622.74,
                    "end": 1622.98,
                    "word": " was",
                    "probability": 0.89990234375
                },
                {
                    "start": 1622.98,
                    "end": 1623.5,
                    "word": " introduced",
                    "probability": 0.74951171875
                },
                {
                    "start": 1623.5,
                    "end": 1623.78,
                    "word": " in",
                    "probability": 0.55517578125
                },
                {
                    "start": 1623.78,
                    "end": 1625.18,
                    "word": " 2018",
                    "probability": 0.6748046875
                },
                {
                    "start": 1625.18,
                    "end": 1626.36,
                    "word": ".",
                    "probability": 0.325927734375
                },
                {
                    "start": 1626.86,
                    "end": 1627.0,
                    "word": " The",
                    "probability": 0.70263671875
                },
                {
                    "start": 1627.0,
                    "end": 1627.38,
                    "word": " model",
                    "probability": 0.85107421875
                },
                {
                    "start": 1627.38,
                    "end": 1628.44,
                    "word": " used",
                    "probability": 0.689453125
                },
                {
                    "start": 1628.44,
                    "end": 1629.62,
                    "word": " bidirectional",
                    "probability": 0.8487955729166666
                },
                {
                    "start": 1629.62,
                    "end": 1630.64,
                    "word": " LSTMs",
                    "probability": 0.9537760416666666
                },
                {
                    "start": 1630.64,
                    "end": 1631.12,
                    "word": " to",
                    "probability": 0.900390625
                }
            ]
        },
        {
            "id": 261,
            "seek": 162574,
            "start": 1631.65,
            "end": 1636.44,
            "text": " generate contextualized word embeddings.",
            "tokens": [
                50868,
                8460,
                35526,
                1602,
                1349,
                12240,
                29432,
                13,
                51044
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3098958263210222,
            "compression_ratio": 1.3006535947712419,
            "no_speech_prob": 0.0023441314697265625,
            "words": [
                {
                    "start": 1631.65,
                    "end": 1632.24,
                    "word": " generate",
                    "probability": 0.80029296875
                },
                {
                    "start": 1632.24,
                    "end": 1633.98,
                    "word": " contextualized",
                    "probability": 0.832275390625
                },
                {
                    "start": 1633.98,
                    "end": 1634.56,
                    "word": " word",
                    "probability": 0.826171875
                },
                {
                    "start": 1634.56,
                    "end": 1635.26,
                    "word": " embeddings",
                    "probability": 0.855224609375
                },
                {
                    "start": 1635.26,
                    "end": 1636.44,
                    "word": ".",
                    "probability": 0.84033203125
                }
            ]
        },
        {
            "id": 262,
            "seek": 162574,
            "start": 1637.65,
            "end": 1647.98,
            "text": " And then, same year later, Google researchers introduced bot and achieved the state-of-art",
            "tokens": [
                51180,
                400,
                550,
                11,
                912,
                1064,
                1780,
                11,
                3329,
                10309,
                7268,
                10592,
                293,
                11042,
                264,
                1785,
                12,
                2670,
                12,
                446,
                51684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3098958263210222,
            "compression_ratio": 1.3006535947712419,
            "no_speech_prob": 0.0023441314697265625,
            "words": [
                {
                    "start": 1637.65,
                    "end": 1638.24,
                    "word": " And",
                    "probability": 0.420166015625
                },
                {
                    "start": 1638.24,
                    "end": 1638.58,
                    "word": " then",
                    "probability": 0.71337890625
                },
                {
                    "start": 1638.58,
                    "end": 1638.8,
                    "word": ",",
                    "probability": 0.58349609375
                },
                {
                    "start": 1638.8,
                    "end": 1639.22,
                    "word": " same",
                    "probability": 0.75634765625
                },
                {
                    "start": 1639.22,
                    "end": 1639.52,
                    "word": " year",
                    "probability": 0.88720703125
                },
                {
                    "start": 1639.52,
                    "end": 1640.04,
                    "word": " later",
                    "probability": 0.8544921875
                },
                {
                    "start": 1640.04,
                    "end": 1640.7,
                    "word": ",",
                    "probability": 0.958984375
                },
                {
                    "start": 1640.7,
                    "end": 1640.94,
                    "word": " Google",
                    "probability": 0.8232421875
                },
                {
                    "start": 1640.94,
                    "end": 1641.74,
                    "word": " researchers",
                    "probability": 0.75146484375
                },
                {
                    "start": 1641.74,
                    "end": 1642.66,
                    "word": " introduced",
                    "probability": 0.7421875
                },
                {
                    "start": 1642.66,
                    "end": 1643.1,
                    "word": " bot",
                    "probability": 0.06536865234375
                },
                {
                    "start": 1643.1,
                    "end": 1644.56,
                    "word": " and",
                    "probability": 0.685546875
                },
                {
                    "start": 1644.56,
                    "end": 1646.28,
                    "word": " achieved",
                    "probability": 0.59716796875
                },
                {
                    "start": 1646.28,
                    "end": 1646.88,
                    "word": " the",
                    "probability": 0.54150390625
                },
                {
                    "start": 1646.88,
                    "end": 1647.32,
                    "word": " state",
                    "probability": 0.8232421875
                },
                {
                    "start": 1647.32,
                    "end": 1647.68,
                    "word": "-of",
                    "probability": 0.88330078125
                },
                {
                    "start": 1647.68,
                    "end": 1647.98,
                    "word": "-art",
                    "probability": 0.827880859375
                }
            ]
        },
        {
            "id": 263,
            "seek": 165398,
            "start": 1647.98,
            "end": 1657.12,
            "text": " performance in many English-understanding problems. Next year, 2019, Baidu researchers",
            "tokens": [
                50364,
                3389,
                294,
                867,
                3669,
                12,
                6617,
                8618,
                2740,
                13,
                3087,
                1064,
                11,
                6071,
                11,
                6777,
                327,
                84,
                10309,
                50828
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24836188314422483,
            "compression_ratio": 1.4915254237288136,
            "no_speech_prob": 0.0195465087890625,
            "words": [
                {
                    "start": 1647.98,
                    "end": 1648.5,
                    "word": " performance",
                    "probability": 0.5830078125
                },
                {
                    "start": 1648.5,
                    "end": 1648.96,
                    "word": " in",
                    "probability": 0.89501953125
                },
                {
                    "start": 1648.96,
                    "end": 1649.32,
                    "word": " many",
                    "probability": 0.8115234375
                },
                {
                    "start": 1649.32,
                    "end": 1650.54,
                    "word": " English",
                    "probability": 0.84619140625
                },
                {
                    "start": 1650.54,
                    "end": 1651.24,
                    "word": "-understanding",
                    "probability": 0.7101236979166666
                },
                {
                    "start": 1651.24,
                    "end": 1652.22,
                    "word": " problems",
                    "probability": 0.6640625
                },
                {
                    "start": 1652.22,
                    "end": 1652.8,
                    "word": ".",
                    "probability": 0.94970703125
                },
                {
                    "start": 1652.8,
                    "end": 1653.42,
                    "word": " Next",
                    "probability": 0.82666015625
                },
                {
                    "start": 1653.42,
                    "end": 1653.74,
                    "word": " year",
                    "probability": 0.8857421875
                },
                {
                    "start": 1653.74,
                    "end": 1653.82,
                    "word": ",",
                    "probability": 0.497802734375
                },
                {
                    "start": 1653.82,
                    "end": 1655.08,
                    "word": " 2019",
                    "probability": 0.6572265625
                },
                {
                    "start": 1655.08,
                    "end": 1656.22,
                    "word": ",",
                    "probability": 0.92919921875
                },
                {
                    "start": 1656.22,
                    "end": 1656.56,
                    "word": " Baidu",
                    "probability": 0.8658854166666666
                },
                {
                    "start": 1656.56,
                    "end": 1657.12,
                    "word": " researchers",
                    "probability": 0.814453125
                }
            ]
        },
        {
            "id": 264,
            "seek": 165398,
            "start": 1657.12,
            "end": 1666.66,
            "text": " introduced the Orni, and the model achieved another state-of-art performance in many",
            "tokens": [
                50828,
                7268,
                264,
                1610,
                3722,
                11,
                293,
                264,
                2316,
                11042,
                1071,
                1785,
                12,
                2670,
                12,
                446,
                3389,
                294,
                867,
                51296
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24836188314422483,
            "compression_ratio": 1.4915254237288136,
            "no_speech_prob": 0.0195465087890625,
            "words": [
                {
                    "start": 1657.12,
                    "end": 1657.84,
                    "word": " introduced",
                    "probability": 0.70654296875
                },
                {
                    "start": 1657.84,
                    "end": 1658.08,
                    "word": " the",
                    "probability": 0.432373046875
                },
                {
                    "start": 1658.08,
                    "end": 1658.44,
                    "word": " Orni",
                    "probability": 0.333953857421875
                },
                {
                    "start": 1658.44,
                    "end": 1659.18,
                    "word": ",",
                    "probability": 0.4736328125
                },
                {
                    "start": 1659.18,
                    "end": 1659.52,
                    "word": " and",
                    "probability": 0.85400390625
                },
                {
                    "start": 1659.52,
                    "end": 1659.86,
                    "word": " the",
                    "probability": 0.79248046875
                },
                {
                    "start": 1659.86,
                    "end": 1660.22,
                    "word": " model",
                    "probability": 0.75
                },
                {
                    "start": 1660.22,
                    "end": 1662.42,
                    "word": " achieved",
                    "probability": 0.7119140625
                },
                {
                    "start": 1662.42,
                    "end": 1663.18,
                    "word": " another",
                    "probability": 0.833984375
                },
                {
                    "start": 1663.18,
                    "end": 1664.04,
                    "word": " state",
                    "probability": 0.8125
                },
                {
                    "start": 1664.04,
                    "end": 1664.4,
                    "word": "-of",
                    "probability": 0.5755615234375
                },
                {
                    "start": 1664.4,
                    "end": 1664.72,
                    "word": "-art",
                    "probability": 0.809814453125
                },
                {
                    "start": 1664.72,
                    "end": 1665.34,
                    "word": " performance",
                    "probability": 0.84814453125
                },
                {
                    "start": 1665.34,
                    "end": 1665.8,
                    "word": " in",
                    "probability": 0.89404296875
                },
                {
                    "start": 1665.8,
                    "end": 1666.66,
                    "word": " many",
                    "probability": 0.81787109375
                }
            ]
        },
        {
            "id": 265,
            "seek": 165398,
            "start": 1666.66,
            "end": 1676.22,
            "text": " Chinese-language-understanding problems. They are all popular characters from Sesame Street.",
            "tokens": [
                51296,
                4649,
                12,
                25241,
                20473,
                12,
                6617,
                8618,
                2740,
                13,
                814,
                366,
                439,
                3743,
                4342,
                490,
                47686,
                7638,
                13,
                51720
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24836188314422483,
            "compression_ratio": 1.4915254237288136,
            "no_speech_prob": 0.0195465087890625,
            "words": [
                {
                    "start": 1666.66,
                    "end": 1667.1,
                    "word": " Chinese",
                    "probability": 0.79833984375
                },
                {
                    "start": 1667.1,
                    "end": 1667.72,
                    "word": "-language",
                    "probability": 0.8212890625
                },
                {
                    "start": 1667.72,
                    "end": 1668.48,
                    "word": "-understanding",
                    "probability": 0.7900390625
                },
                {
                    "start": 1668.48,
                    "end": 1669.22,
                    "word": " problems",
                    "probability": 0.70556640625
                },
                {
                    "start": 1669.22,
                    "end": 1670.34,
                    "word": ".",
                    "probability": 0.97607421875
                },
                {
                    "start": 1671.14,
                    "end": 1671.36,
                    "word": " They",
                    "probability": 0.68310546875
                },
                {
                    "start": 1671.36,
                    "end": 1671.58,
                    "word": " are",
                    "probability": 0.83740234375
                },
                {
                    "start": 1671.58,
                    "end": 1671.9,
                    "word": " all",
                    "probability": 0.87158203125
                },
                {
                    "start": 1671.9,
                    "end": 1672.52,
                    "word": " popular",
                    "probability": 0.93798828125
                },
                {
                    "start": 1672.52,
                    "end": 1673.78,
                    "word": " characters",
                    "probability": 0.7958984375
                },
                {
                    "start": 1673.78,
                    "end": 1674.22,
                    "word": " from",
                    "probability": 0.8134765625
                },
                {
                    "start": 1674.22,
                    "end": 1674.72,
                    "word": " Sesame",
                    "probability": 0.94873046875
                },
                {
                    "start": 1674.72,
                    "end": 1675.14,
                    "word": " Street",
                    "probability": 0.720703125
                },
                {
                    "start": 1675.14,
                    "end": 1676.22,
                    "word": ".",
                    "probability": 0.9326171875
                }
            ]
        },
        {
            "id": 266,
            "seek": 168228,
            "start": 1677.18,
            "end": 1686.88,
            "text": " This year, 2020, we introduced CatBot to tackle email security problems.",
            "tokens": [
                50412,
                639,
                1064,
                11,
                4808,
                11,
                321,
                7268,
                9565,
                33,
                310,
                281,
                14896,
                3796,
                3825,
                2740,
                13,
                50816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2431818122213537,
            "compression_ratio": 1.2941176470588236,
            "no_speech_prob": 0.0017375946044921875,
            "words": [
                {
                    "start": 1677.18,
                    "end": 1677.84,
                    "word": " This",
                    "probability": 0.751953125
                },
                {
                    "start": 1677.84,
                    "end": 1678.16,
                    "word": " year",
                    "probability": 0.84326171875
                },
                {
                    "start": 1678.16,
                    "end": 1678.26,
                    "word": ",",
                    "probability": 0.39453125
                },
                {
                    "start": 1678.26,
                    "end": 1679.04,
                    "word": " 2020",
                    "probability": 0.7333984375
                },
                {
                    "start": 1679.04,
                    "end": 1679.64,
                    "word": ",",
                    "probability": 0.87451171875
                },
                {
                    "start": 1679.64,
                    "end": 1679.8,
                    "word": " we",
                    "probability": 0.8984375
                },
                {
                    "start": 1679.8,
                    "end": 1680.42,
                    "word": " introduced",
                    "probability": 0.6396484375
                },
                {
                    "start": 1680.42,
                    "end": 1681.22,
                    "word": " CatBot",
                    "probability": 0.7556966145833334
                },
                {
                    "start": 1681.22,
                    "end": 1682.12,
                    "word": " to",
                    "probability": 0.88525390625
                },
                {
                    "start": 1682.12,
                    "end": 1682.72,
                    "word": " tackle",
                    "probability": 0.76904296875
                },
                {
                    "start": 1682.72,
                    "end": 1684.14,
                    "word": " email",
                    "probability": 0.76708984375
                },
                {
                    "start": 1684.14,
                    "end": 1684.74,
                    "word": " security",
                    "probability": 0.890625
                },
                {
                    "start": 1684.74,
                    "end": 1685.68,
                    "word": " problems",
                    "probability": 0.68603515625
                },
                {
                    "start": 1685.68,
                    "end": 1686.88,
                    "word": ".",
                    "probability": 0.95458984375
                }
            ]
        },
        {
            "id": 267,
            "seek": 168228,
            "start": 1692.01,
            "end": 1700.27,
            "text": " Transformer-based NLP models are powerful, but they are complex and heavy.",
            "tokens": [
                50908,
                27938,
                260,
                12,
                6032,
                426,
                45196,
                5245,
                366,
                4005,
                11,
                457,
                436,
                366,
                3997,
                293,
                4676,
                13,
                51276
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2431818122213537,
            "compression_ratio": 1.2941176470588236,
            "no_speech_prob": 0.0017375946044921875,
            "words": [
                {
                    "start": 1692.01,
                    "end": 1692.79,
                    "word": " Transformer",
                    "probability": 0.830078125
                },
                {
                    "start": 1692.79,
                    "end": 1693.15,
                    "word": "-based",
                    "probability": 0.786376953125
                },
                {
                    "start": 1693.15,
                    "end": 1693.95,
                    "word": " NLP",
                    "probability": 0.972900390625
                },
                {
                    "start": 1693.95,
                    "end": 1694.53,
                    "word": " models",
                    "probability": 0.81396484375
                },
                {
                    "start": 1694.53,
                    "end": 1695.01,
                    "word": " are",
                    "probability": 0.89599609375
                },
                {
                    "start": 1695.01,
                    "end": 1695.57,
                    "word": " powerful",
                    "probability": 0.8212890625
                },
                {
                    "start": 1695.57,
                    "end": 1696.05,
                    "word": ",",
                    "probability": 0.56982421875
                },
                {
                    "start": 1696.05,
                    "end": 1696.17,
                    "word": " but",
                    "probability": 0.8994140625
                },
                {
                    "start": 1696.17,
                    "end": 1696.49,
                    "word": " they",
                    "probability": 0.69775390625
                },
                {
                    "start": 1696.49,
                    "end": 1696.73,
                    "word": " are",
                    "probability": 0.8525390625
                },
                {
                    "start": 1696.73,
                    "end": 1697.39,
                    "word": " complex",
                    "probability": 0.71728515625
                },
                {
                    "start": 1697.39,
                    "end": 1697.95,
                    "word": " and",
                    "probability": 0.88134765625
                },
                {
                    "start": 1697.95,
                    "end": 1699.23,
                    "word": " heavy",
                    "probability": 0.734375
                },
                {
                    "start": 1699.23,
                    "end": 1700.27,
                    "word": ".",
                    "probability": 0.7802734375
                }
            ]
        },
        {
            "id": 268,
            "seek": 168228,
            "start": 1700.79,
            "end": 1709.05,
            "text": " And it is challenging to deploy heavy models for real-time applications.",
            "tokens": [
                51368,
                400,
                309,
                307,
                7595,
                281,
                7274,
                4676,
                5245,
                337,
                957,
                12,
                3766,
                5821,
                13,
                51724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2431818122213537,
            "compression_ratio": 1.2941176470588236,
            "no_speech_prob": 0.0017375946044921875,
            "words": [
                {
                    "start": 1700.79,
                    "end": 1701.39,
                    "word": " And",
                    "probability": 0.6923828125
                },
                {
                    "start": 1701.39,
                    "end": 1701.67,
                    "word": " it",
                    "probability": 0.83056640625
                },
                {
                    "start": 1701.67,
                    "end": 1701.83,
                    "word": " is",
                    "probability": 0.89404296875
                },
                {
                    "start": 1701.83,
                    "end": 1702.39,
                    "word": " challenging",
                    "probability": 0.8251953125
                },
                {
                    "start": 1702.39,
                    "end": 1702.81,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 1702.81,
                    "end": 1703.35,
                    "word": " deploy",
                    "probability": 0.7783203125
                },
                {
                    "start": 1703.35,
                    "end": 1704.09,
                    "word": " heavy",
                    "probability": 0.67919921875
                },
                {
                    "start": 1704.09,
                    "end": 1704.71,
                    "word": " models",
                    "probability": 0.8046875
                },
                {
                    "start": 1704.71,
                    "end": 1705.55,
                    "word": " for",
                    "probability": 0.810546875
                },
                {
                    "start": 1705.55,
                    "end": 1706.61,
                    "word": " real",
                    "probability": 0.8603515625
                },
                {
                    "start": 1706.61,
                    "end": 1707.01,
                    "word": "-time",
                    "probability": 0.840576171875
                },
                {
                    "start": 1707.01,
                    "end": 1708.11,
                    "word": " applications",
                    "probability": 0.76806640625
                },
                {
                    "start": 1708.11,
                    "end": 1709.05,
                    "word": ".",
                    "probability": 0.93994140625
                }
            ]
        },
        {
            "id": 269,
            "seek": 170712,
            "start": 1709.67,
            "end": 1718.77,
            "text": " So our first design goal is to convert the heavy model into a lightweight model,",
            "tokens": [
                50400,
                407,
                527,
                700,
                1715,
                3387,
                307,
                281,
                7620,
                264,
                4676,
                2316,
                666,
                257,
                22052,
                2316,
                11,
                50836
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25727372128387976,
            "compression_ratio": 1.4137931034482758,
            "no_speech_prob": 0.004108428955078125,
            "words": [
                {
                    "start": 1709.67,
                    "end": 1710.11,
                    "word": " So",
                    "probability": 0.260498046875
                },
                {
                    "start": 1710.11,
                    "end": 1710.53,
                    "word": " our",
                    "probability": 0.6123046875
                },
                {
                    "start": 1710.53,
                    "end": 1711.33,
                    "word": " first",
                    "probability": 0.76611328125
                },
                {
                    "start": 1711.33,
                    "end": 1711.75,
                    "word": " design",
                    "probability": 0.90380859375
                },
                {
                    "start": 1711.75,
                    "end": 1712.09,
                    "word": " goal",
                    "probability": 0.8779296875
                },
                {
                    "start": 1712.09,
                    "end": 1712.41,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 1712.41,
                    "end": 1712.83,
                    "word": " to",
                    "probability": 0.89599609375
                },
                {
                    "start": 1712.83,
                    "end": 1713.41,
                    "word": " convert",
                    "probability": 0.85546875
                },
                {
                    "start": 1713.41,
                    "end": 1714.39,
                    "word": " the",
                    "probability": 0.70703125
                },
                {
                    "start": 1714.39,
                    "end": 1714.81,
                    "word": " heavy",
                    "probability": 0.67333984375
                },
                {
                    "start": 1714.81,
                    "end": 1716.49,
                    "word": " model",
                    "probability": 0.697265625
                },
                {
                    "start": 1716.49,
                    "end": 1717.15,
                    "word": " into",
                    "probability": 0.7099609375
                },
                {
                    "start": 1717.53,
                    "end": 1718.03,
                    "word": " a lightweight",
                    "probability": 0.78271484375
                },
                {
                    "start": 1718.03,
                    "end": 1718.59,
                    "word": " model",
                    "probability": 0.81689453125
                },
                {
                    "start": 1718.59,
                    "end": 1718.77,
                    "word": ",",
                    "probability": 0.2410888671875
                }
            ]
        },
        {
            "id": 270,
            "seek": 170712,
            "start": 1718.77,
            "end": 1725.81,
            "text": " so we can reduce the number of parameters and then we can improve inference speed.",
            "tokens": [
                50868,
                370,
                321,
                393,
                5407,
                264,
                1230,
                295,
                9834,
                293,
                550,
                321,
                393,
                3470,
                38253,
                3073,
                13,
                51156
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25727372128387976,
            "compression_ratio": 1.4137931034482758,
            "no_speech_prob": 0.004108428955078125,
            "words": [
                {
                    "start": 1718.77,
                    "end": 1719.41,
                    "word": " so",
                    "probability": 0.84814453125
                },
                {
                    "start": 1719.41,
                    "end": 1719.61,
                    "word": " we",
                    "probability": 0.8828125
                },
                {
                    "start": 1719.61,
                    "end": 1719.87,
                    "word": " can",
                    "probability": 0.89892578125
                },
                {
                    "start": 1719.87,
                    "end": 1720.31,
                    "word": " reduce",
                    "probability": 0.76318359375
                },
                {
                    "start": 1720.31,
                    "end": 1720.63,
                    "word": " the",
                    "probability": 0.469482421875
                },
                {
                    "start": 1720.63,
                    "end": 1720.83,
                    "word": " number",
                    "probability": 0.8701171875
                },
                {
                    "start": 1720.83,
                    "end": 1721.03,
                    "word": " of",
                    "probability": 0.90966796875
                },
                {
                    "start": 1721.03,
                    "end": 1721.57,
                    "word": " parameters",
                    "probability": 0.88427734375
                },
                {
                    "start": 1721.57,
                    "end": 1722.11,
                    "word": " and",
                    "probability": 0.70068359375
                },
                {
                    "start": 1722.11,
                    "end": 1722.33,
                    "word": " then",
                    "probability": 0.60107421875
                },
                {
                    "start": 1722.33,
                    "end": 1722.55,
                    "word": " we",
                    "probability": 0.8349609375
                },
                {
                    "start": 1722.55,
                    "end": 1722.87,
                    "word": " can",
                    "probability": 0.88671875
                },
                {
                    "start": 1722.87,
                    "end": 1723.77,
                    "word": " improve",
                    "probability": 0.72119140625
                },
                {
                    "start": 1723.77,
                    "end": 1724.53,
                    "word": " inference",
                    "probability": 0.4423828125
                },
                {
                    "start": 1724.53,
                    "end": 1724.93,
                    "word": " speed",
                    "probability": 0.83447265625
                },
                {
                    "start": 1724.93,
                    "end": 1725.81,
                    "word": ".",
                    "probability": 0.93017578125
                }
            ]
        },
        {
            "id": 271,
            "seek": 170712,
            "start": 1726.45,
            "end": 1733.89,
            "text": " We downsized a baseline model called DistribBot, which has six transformer blocks.",
            "tokens": [
                51240,
                492,
                760,
                20614,
                257,
                20518,
                2316,
                1219,
                9840,
                2024,
                33,
                310,
                11,
                597,
                575,
                2309,
                31782,
                8474,
                13,
                51560
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25727372128387976,
            "compression_ratio": 1.4137931034482758,
            "no_speech_prob": 0.004108428955078125,
            "words": [
                {
                    "start": 1726.45,
                    "end": 1726.89,
                    "word": " We",
                    "probability": 0.89697265625
                },
                {
                    "start": 1726.89,
                    "end": 1727.75,
                    "word": " downsized",
                    "probability": 0.6802978515625
                },
                {
                    "start": 1728.11,
                    "end": 1728.53,
                    "word": " a baseline",
                    "probability": 0.8173828125
                },
                {
                    "start": 1728.53,
                    "end": 1729.13,
                    "word": " model",
                    "probability": 0.84912109375
                },
                {
                    "start": 1729.13,
                    "end": 1729.67,
                    "word": " called",
                    "probability": 0.75830078125
                },
                {
                    "start": 1729.67,
                    "end": 1730.53,
                    "word": " DistribBot",
                    "probability": 0.596923828125
                },
                {
                    "start": 1730.53,
                    "end": 1730.95,
                    "word": ",",
                    "probability": 0.5966796875
                },
                {
                    "start": 1730.95,
                    "end": 1731.23,
                    "word": " which",
                    "probability": 0.9169921875
                },
                {
                    "start": 1731.23,
                    "end": 1731.49,
                    "word": " has",
                    "probability": 0.8916015625
                },
                {
                    "start": 1731.49,
                    "end": 1731.95,
                    "word": " six",
                    "probability": 0.428955078125
                },
                {
                    "start": 1731.95,
                    "end": 1732.57,
                    "word": " transformer",
                    "probability": 0.68212890625
                },
                {
                    "start": 1732.57,
                    "end": 1733.01,
                    "word": " blocks",
                    "probability": 0.83203125
                },
                {
                    "start": 1733.01,
                    "end": 1733.89,
                    "word": ".",
                    "probability": 0.9609375
                }
            ]
        },
        {
            "id": 272,
            "seek": 173174,
            "start": 1734.59,
            "end": 1740.65,
            "text": " We take half of transformer blocks from pre-trained model and then replace missing",
            "tokens": [
                50404,
                492,
                747,
                1922,
                295,
                31782,
                8474,
                490,
                659,
                12,
                17227,
                2001,
                2316,
                293,
                550,
                7406,
                5361,
                50716
            ],
            "temperature": 0.0,
            "avg_logprob": -0.256200409124768,
            "compression_ratio": 1.6129032258064515,
            "no_speech_prob": 0.005252838134765625,
            "words": [
                {
                    "start": 1734.59,
                    "end": 1734.99,
                    "word": " We",
                    "probability": 0.77001953125
                },
                {
                    "start": 1734.99,
                    "end": 1735.39,
                    "word": " take",
                    "probability": 0.74755859375
                },
                {
                    "start": 1735.39,
                    "end": 1735.89,
                    "word": " half",
                    "probability": 0.7978515625
                },
                {
                    "start": 1735.89,
                    "end": 1736.13,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1736.13,
                    "end": 1736.63,
                    "word": " transformer",
                    "probability": 0.350341796875
                },
                {
                    "start": 1736.63,
                    "end": 1737.03,
                    "word": " blocks",
                    "probability": 0.82373046875
                },
                {
                    "start": 1737.03,
                    "end": 1737.35,
                    "word": " from",
                    "probability": 0.8017578125
                },
                {
                    "start": 1737.35,
                    "end": 1737.69,
                    "word": " pre",
                    "probability": 0.69189453125
                },
                {
                    "start": 1737.69,
                    "end": 1738.01,
                    "word": "-trained",
                    "probability": 0.8992513020833334
                },
                {
                    "start": 1738.01,
                    "end": 1738.41,
                    "word": " model",
                    "probability": 0.63916015625
                },
                {
                    "start": 1738.41,
                    "end": 1739.13,
                    "word": " and",
                    "probability": 0.52001953125
                },
                {
                    "start": 1739.13,
                    "end": 1739.37,
                    "word": " then",
                    "probability": 0.6767578125
                },
                {
                    "start": 1739.37,
                    "end": 1740.01,
                    "word": " replace",
                    "probability": 0.80078125
                },
                {
                    "start": 1740.01,
                    "end": 1740.65,
                    "word": " missing",
                    "probability": 0.80615234375
                }
            ]
        },
        {
            "id": 273,
            "seek": 173174,
            "start": 1740.65,
            "end": 1750.11,
            "text": " transformers with simple adapters. For example here, we take transformer 1, 3, 5,",
            "tokens": [
                50716,
                4088,
                433,
                365,
                2199,
                23169,
                1559,
                13,
                1171,
                1365,
                510,
                11,
                321,
                747,
                31782,
                502,
                11,
                805,
                11,
                1025,
                11,
                51144
            ],
            "temperature": 0.0,
            "avg_logprob": -0.256200409124768,
            "compression_ratio": 1.6129032258064515,
            "no_speech_prob": 0.005252838134765625,
            "words": [
                {
                    "start": 1740.65,
                    "end": 1741.75,
                    "word": " transformers",
                    "probability": 0.899169921875
                },
                {
                    "start": 1741.75,
                    "end": 1742.01,
                    "word": " with",
                    "probability": 0.82666015625
                },
                {
                    "start": 1742.01,
                    "end": 1742.81,
                    "word": " simple",
                    "probability": 0.8505859375
                },
                {
                    "start": 1742.81,
                    "end": 1743.49,
                    "word": " adapters",
                    "probability": 0.844482421875
                },
                {
                    "start": 1743.49,
                    "end": 1744.29,
                    "word": ".",
                    "probability": 0.93017578125
                },
                {
                    "start": 1744.61,
                    "end": 1744.79,
                    "word": " For",
                    "probability": 0.8466796875
                },
                {
                    "start": 1744.79,
                    "end": 1745.59,
                    "word": " example",
                    "probability": 0.91357421875
                },
                {
                    "start": 1745.59,
                    "end": 1746.15,
                    "word": " here",
                    "probability": 0.4609375
                },
                {
                    "start": 1746.15,
                    "end": 1746.37,
                    "word": ",",
                    "probability": 0.81884765625
                },
                {
                    "start": 1746.37,
                    "end": 1746.53,
                    "word": " we",
                    "probability": 0.87890625
                },
                {
                    "start": 1746.53,
                    "end": 1746.99,
                    "word": " take",
                    "probability": 0.77490234375
                },
                {
                    "start": 1746.99,
                    "end": 1748.15,
                    "word": " transformer",
                    "probability": 0.72900390625
                },
                {
                    "start": 1748.15,
                    "end": 1748.61,
                    "word": " 1",
                    "probability": 0.83740234375
                },
                {
                    "start": 1748.61,
                    "end": 1748.83,
                    "word": ",",
                    "probability": 0.8271484375
                },
                {
                    "start": 1748.83,
                    "end": 1749.11,
                    "word": " 3",
                    "probability": 0.73681640625
                },
                {
                    "start": 1749.11,
                    "end": 1749.29,
                    "word": ",",
                    "probability": 0.9345703125
                },
                {
                    "start": 1749.29,
                    "end": 1749.61,
                    "word": " 5",
                    "probability": 0.95166015625
                },
                {
                    "start": 1749.61,
                    "end": 1750.11,
                    "word": ",",
                    "probability": 0.361328125
                }
            ]
        },
        {
            "id": 274,
            "seek": 173174,
            "start": 1750.11,
            "end": 1758.51,
            "text": " and then we added two adapters. Also, we can take other number of transformer blocks.",
            "tokens": [
                51176,
                293,
                550,
                321,
                3869,
                732,
                23169,
                1559,
                13,
                2743,
                11,
                321,
                393,
                747,
                661,
                1230,
                295,
                31782,
                8474,
                13,
                51552
            ],
            "temperature": 0.0,
            "avg_logprob": -0.256200409124768,
            "compression_ratio": 1.6129032258064515,
            "no_speech_prob": 0.005252838134765625,
            "words": [
                {
                    "start": 1750.11,
                    "end": 1750.45,
                    "word": " and",
                    "probability": 0.8447265625
                },
                {
                    "start": 1750.45,
                    "end": 1750.65,
                    "word": " then",
                    "probability": 0.72265625
                },
                {
                    "start": 1750.65,
                    "end": 1750.85,
                    "word": " we",
                    "probability": 0.83203125
                },
                {
                    "start": 1750.85,
                    "end": 1751.35,
                    "word": " added",
                    "probability": 0.77099609375
                },
                {
                    "start": 1751.35,
                    "end": 1752.49,
                    "word": " two",
                    "probability": 0.54736328125
                },
                {
                    "start": 1752.49,
                    "end": 1753.25,
                    "word": " adapters",
                    "probability": 0.8876953125
                },
                {
                    "start": 1753.25,
                    "end": 1754.05,
                    "word": ".",
                    "probability": 0.9677734375
                },
                {
                    "start": 1754.51,
                    "end": 1754.83,
                    "word": " Also",
                    "probability": 0.86376953125
                },
                {
                    "start": 1754.83,
                    "end": 1754.95,
                    "word": ",",
                    "probability": 0.572265625
                },
                {
                    "start": 1754.95,
                    "end": 1754.99,
                    "word": " we",
                    "probability": 0.89794921875
                },
                {
                    "start": 1754.99,
                    "end": 1755.13,
                    "word": " can",
                    "probability": 0.8857421875
                },
                {
                    "start": 1755.13,
                    "end": 1755.51,
                    "word": " take",
                    "probability": 0.80322265625
                },
                {
                    "start": 1755.51,
                    "end": 1756.21,
                    "word": " other",
                    "probability": 0.5234375
                },
                {
                    "start": 1756.21,
                    "end": 1756.55,
                    "word": " number",
                    "probability": 0.8271484375
                },
                {
                    "start": 1756.55,
                    "end": 1756.81,
                    "word": " of",
                    "probability": 0.912109375
                },
                {
                    "start": 1756.81,
                    "end": 1757.23,
                    "word": " transformer",
                    "probability": 0.70703125
                },
                {
                    "start": 1757.23,
                    "end": 1757.71,
                    "word": " blocks",
                    "probability": 0.8232421875
                },
                {
                    "start": 1757.71,
                    "end": 1758.51,
                    "word": ".",
                    "probability": 0.9677734375
                }
            ]
        },
        {
            "id": 275,
            "seek": 175972,
            "start": 1759.32,
            "end": 1764.84,
            "text": " This approach allows simply significantly reduced number of parameters.",
            "tokens": [
                50412,
                639,
                3109,
                4045,
                2935,
                10591,
                9212,
                1230,
                295,
                9834,
                13,
                50652
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23469387268533512,
            "compression_ratio": 1.409356725146199,
            "no_speech_prob": 0.005046844482421875,
            "words": [
                {
                    "start": 1759.32,
                    "end": 1759.91,
                    "word": " This",
                    "probability": 0.7470703125
                },
                {
                    "start": 1759.91,
                    "end": 1760.41,
                    "word": " approach",
                    "probability": 0.8095703125
                },
                {
                    "start": 1760.41,
                    "end": 1760.91,
                    "word": " allows",
                    "probability": 0.67236328125
                },
                {
                    "start": 1760.91,
                    "end": 1761.57,
                    "word": " simply",
                    "probability": 0.40185546875
                },
                {
                    "start": 1761.57,
                    "end": 1762.51,
                    "word": " significantly",
                    "probability": 0.8076171875
                },
                {
                    "start": 1762.51,
                    "end": 1763.09,
                    "word": " reduced",
                    "probability": 0.7646484375
                },
                {
                    "start": 1763.09,
                    "end": 1763.37,
                    "word": " number",
                    "probability": 0.837890625
                },
                {
                    "start": 1763.37,
                    "end": 1763.59,
                    "word": " of",
                    "probability": 0.90966796875
                },
                {
                    "start": 1763.59,
                    "end": 1764.25,
                    "word": " parameters",
                    "probability": 0.86083984375
                },
                {
                    "start": 1764.25,
                    "end": 1764.84,
                    "word": ".",
                    "probability": 0.91162109375
                }
            ]
        },
        {
            "id": 276,
            "seek": 175972,
            "start": 1769.94,
            "end": 1779.02,
            "text": " The second goal is to improve the model performance by combining additional input.",
            "tokens": [
                50748,
                440,
                1150,
                3387,
                307,
                281,
                3470,
                264,
                2316,
                3389,
                538,
                21928,
                4497,
                4846,
                13,
                51144
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23469387268533512,
            "compression_ratio": 1.409356725146199,
            "no_speech_prob": 0.005046844482421875,
            "words": [
                {
                    "start": 1769.94,
                    "end": 1770.16,
                    "word": " The",
                    "probability": 0.7880859375
                },
                {
                    "start": 1770.16,
                    "end": 1770.54,
                    "word": " second",
                    "probability": 0.787109375
                },
                {
                    "start": 1770.54,
                    "end": 1771.18,
                    "word": " goal",
                    "probability": 0.89990234375
                },
                {
                    "start": 1771.18,
                    "end": 1771.58,
                    "word": " is",
                    "probability": 0.9033203125
                },
                {
                    "start": 1771.58,
                    "end": 1771.98,
                    "word": " to",
                    "probability": 0.89794921875
                },
                {
                    "start": 1771.98,
                    "end": 1772.58,
                    "word": " improve",
                    "probability": 0.728515625
                },
                {
                    "start": 1772.58,
                    "end": 1773.08,
                    "word": " the",
                    "probability": 0.6611328125
                },
                {
                    "start": 1773.08,
                    "end": 1773.44,
                    "word": " model",
                    "probability": 0.84619140625
                },
                {
                    "start": 1773.44,
                    "end": 1774.12,
                    "word": " performance",
                    "probability": 0.81591796875
                },
                {
                    "start": 1774.12,
                    "end": 1774.86,
                    "word": " by",
                    "probability": 0.89013671875
                },
                {
                    "start": 1774.86,
                    "end": 1775.64,
                    "word": " combining",
                    "probability": 0.83447265625
                },
                {
                    "start": 1775.64,
                    "end": 1777.18,
                    "word": " additional",
                    "probability": 0.80078125
                },
                {
                    "start": 1777.18,
                    "end": 1777.84,
                    "word": " input",
                    "probability": 0.7470703125
                },
                {
                    "start": 1777.84,
                    "end": 1779.02,
                    "word": ".",
                    "probability": 0.9658203125
                }
            ]
        },
        {
            "id": 277,
            "seek": 175972,
            "start": 1780.42,
            "end": 1790.16,
            "text": " Standard NLP models only accept text data as input. However, we can extract additional",
            "tokens": [
                51292,
                21298,
                426,
                45196,
                5245,
                787,
                3241,
                2487,
                1412,
                382,
                4846,
                13,
                2908,
                11,
                321,
                393,
                8947,
                4497,
                51752
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23469387268533512,
            "compression_ratio": 1.409356725146199,
            "no_speech_prob": 0.005046844482421875,
            "words": [
                {
                    "start": 1780.42,
                    "end": 1781.38,
                    "word": " Standard",
                    "probability": 0.8076171875
                },
                {
                    "start": 1781.38,
                    "end": 1781.92,
                    "word": " NLP",
                    "probability": 0.9658203125
                },
                {
                    "start": 1781.92,
                    "end": 1782.36,
                    "word": " models",
                    "probability": 0.8232421875
                },
                {
                    "start": 1782.36,
                    "end": 1782.88,
                    "word": " only",
                    "probability": 0.79833984375
                },
                {
                    "start": 1782.88,
                    "end": 1784.16,
                    "word": " accept",
                    "probability": 0.72021484375
                },
                {
                    "start": 1784.16,
                    "end": 1784.72,
                    "word": " text",
                    "probability": 0.82177734375
                },
                {
                    "start": 1784.72,
                    "end": 1785.0,
                    "word": " data",
                    "probability": 0.84228515625
                },
                {
                    "start": 1785.0,
                    "end": 1785.26,
                    "word": " as",
                    "probability": 0.8955078125
                },
                {
                    "start": 1785.26,
                    "end": 1785.64,
                    "word": " input",
                    "probability": 0.7880859375
                },
                {
                    "start": 1785.64,
                    "end": 1786.66,
                    "word": ".",
                    "probability": 0.732421875
                },
                {
                    "start": 1786.66,
                    "end": 1787.34,
                    "word": " However",
                    "probability": 0.857421875
                },
                {
                    "start": 1787.34,
                    "end": 1788.12,
                    "word": ",",
                    "probability": 0.943359375
                },
                {
                    "start": 1788.12,
                    "end": 1788.4,
                    "word": " we",
                    "probability": 0.90234375
                },
                {
                    "start": 1788.4,
                    "end": 1788.7,
                    "word": " can",
                    "probability": 0.888671875
                },
                {
                    "start": 1788.7,
                    "end": 1789.32,
                    "word": " extract",
                    "probability": 0.7099609375
                },
                {
                    "start": 1789.32,
                    "end": 1790.16,
                    "word": " additional",
                    "probability": 0.8447265625
                }
            ]
        },
        {
            "id": 278,
            "seek": 178266,
            "start": 1790.75,
            "end": 1798.42,
            "text": " features from email headers and we can use the additional input to our model.",
            "tokens": [
                50396,
                4122,
                490,
                3796,
                45101,
                293,
                321,
                393,
                764,
                264,
                4497,
                4846,
                281,
                527,
                2316,
                13,
                50732
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23480113311247391,
            "compression_ratio": 1.8345323741007193,
            "no_speech_prob": 0.003459930419921875,
            "words": [
                {
                    "start": 1790.75,
                    "end": 1791.2,
                    "word": " features",
                    "probability": 0.01983642578125
                },
                {
                    "start": 1791.2,
                    "end": 1791.72,
                    "word": " from",
                    "probability": 0.734375
                },
                {
                    "start": 1791.72,
                    "end": 1792.1,
                    "word": " email",
                    "probability": 0.8125
                },
                {
                    "start": 1792.1,
                    "end": 1792.68,
                    "word": " headers",
                    "probability": 0.92431640625
                },
                {
                    "start": 1792.68,
                    "end": 1793.58,
                    "word": " and",
                    "probability": 0.44384765625
                },
                {
                    "start": 1793.58,
                    "end": 1793.82,
                    "word": " we",
                    "probability": 0.87646484375
                },
                {
                    "start": 1793.82,
                    "end": 1794.02,
                    "word": " can",
                    "probability": 0.892578125
                },
                {
                    "start": 1794.02,
                    "end": 1794.32,
                    "word": " use",
                    "probability": 0.80126953125
                },
                {
                    "start": 1794.32,
                    "end": 1794.56,
                    "word": " the",
                    "probability": 0.75390625
                },
                {
                    "start": 1794.56,
                    "end": 1795.0,
                    "word": " additional",
                    "probability": 0.8525390625
                },
                {
                    "start": 1795.0,
                    "end": 1795.74,
                    "word": " input",
                    "probability": 0.85546875
                },
                {
                    "start": 1795.74,
                    "end": 1795.98,
                    "word": " to",
                    "probability": 0.876953125
                },
                {
                    "start": 1795.98,
                    "end": 1797.1,
                    "word": " our",
                    "probability": 0.476806640625
                },
                {
                    "start": 1797.1,
                    "end": 1797.52,
                    "word": " model",
                    "probability": 0.8134765625
                },
                {
                    "start": 1797.52,
                    "end": 1798.42,
                    "word": ".",
                    "probability": 0.92236328125
                }
            ]
        },
        {
            "id": 279,
            "seek": 178266,
            "start": 1798.9,
            "end": 1807.66,
            "text": " So the text input will be the input to the embedding and then we can add additional input",
            "tokens": [
                50808,
                407,
                264,
                2487,
                4846,
                486,
                312,
                264,
                4846,
                281,
                264,
                12240,
                3584,
                293,
                550,
                321,
                393,
                909,
                4497,
                4846,
                51248
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23480113311247391,
            "compression_ratio": 1.8345323741007193,
            "no_speech_prob": 0.003459930419921875,
            "words": [
                {
                    "start": 1798.9,
                    "end": 1799.38,
                    "word": " So",
                    "probability": 0.6337890625
                },
                {
                    "start": 1799.38,
                    "end": 1799.94,
                    "word": " the",
                    "probability": 0.6533203125
                },
                {
                    "start": 1799.94,
                    "end": 1801.22,
                    "word": " text",
                    "probability": 0.86083984375
                },
                {
                    "start": 1801.22,
                    "end": 1801.66,
                    "word": " input",
                    "probability": 0.83740234375
                },
                {
                    "start": 1801.66,
                    "end": 1801.98,
                    "word": " will",
                    "probability": 0.77685546875
                },
                {
                    "start": 1801.98,
                    "end": 1802.08,
                    "word": " be",
                    "probability": 0.8740234375
                },
                {
                    "start": 1802.08,
                    "end": 1802.24,
                    "word": " the",
                    "probability": 0.7529296875
                },
                {
                    "start": 1802.24,
                    "end": 1802.52,
                    "word": " input",
                    "probability": 0.89697265625
                },
                {
                    "start": 1802.52,
                    "end": 1802.72,
                    "word": " to",
                    "probability": 0.892578125
                },
                {
                    "start": 1802.72,
                    "end": 1802.88,
                    "word": " the",
                    "probability": 0.78076171875
                },
                {
                    "start": 1802.88,
                    "end": 1803.54,
                    "word": " embedding",
                    "probability": 0.92236328125
                },
                {
                    "start": 1803.54,
                    "end": 1804.28,
                    "word": " and",
                    "probability": 0.6123046875
                },
                {
                    "start": 1804.28,
                    "end": 1804.56,
                    "word": " then",
                    "probability": 0.72802734375
                },
                {
                    "start": 1804.56,
                    "end": 1805.42,
                    "word": " we",
                    "probability": 0.89013671875
                },
                {
                    "start": 1805.42,
                    "end": 1805.76,
                    "word": " can",
                    "probability": 0.8955078125
                },
                {
                    "start": 1805.76,
                    "end": 1806.66,
                    "word": " add",
                    "probability": 0.82568359375
                },
                {
                    "start": 1806.66,
                    "end": 1807.16,
                    "word": " additional",
                    "probability": 0.82666015625
                },
                {
                    "start": 1807.16,
                    "end": 1807.66,
                    "word": " input",
                    "probability": 0.86962890625
                }
            ]
        },
        {
            "id": 280,
            "seek": 178266,
            "start": 1807.66,
            "end": 1813.1,
            "text": " to the classification header and we added additional dense layers in the classification",
            "tokens": [
                51248,
                281,
                264,
                21538,
                23117,
                293,
                321,
                3869,
                4497,
                18011,
                7914,
                294,
                264,
                21538,
                51524
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23480113311247391,
            "compression_ratio": 1.8345323741007193,
            "no_speech_prob": 0.003459930419921875,
            "words": [
                {
                    "start": 1807.66,
                    "end": 1808.16,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 1808.16,
                    "end": 1808.28,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 1808.28,
                    "end": 1808.94,
                    "word": " classification",
                    "probability": 0.92529296875
                },
                {
                    "start": 1808.94,
                    "end": 1809.4,
                    "word": " header",
                    "probability": 0.2646484375
                },
                {
                    "start": 1809.4,
                    "end": 1809.92,
                    "word": " and",
                    "probability": 0.63427734375
                },
                {
                    "start": 1809.92,
                    "end": 1810.16,
                    "word": " we",
                    "probability": 0.8984375
                },
                {
                    "start": 1810.16,
                    "end": 1810.5,
                    "word": " added",
                    "probability": 0.77294921875
                },
                {
                    "start": 1810.5,
                    "end": 1811.04,
                    "word": " additional",
                    "probability": 0.77880859375
                },
                {
                    "start": 1811.04,
                    "end": 1811.58,
                    "word": " dense",
                    "probability": 0.8525390625
                },
                {
                    "start": 1811.58,
                    "end": 1812.24,
                    "word": " layers",
                    "probability": 0.857421875
                },
                {
                    "start": 1812.24,
                    "end": 1812.52,
                    "word": " in",
                    "probability": 0.9033203125
                },
                {
                    "start": 1812.52,
                    "end": 1812.68,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 1812.68,
                    "end": 1813.1,
                    "word": " classification",
                    "probability": 0.9189453125
                }
            ]
        },
        {
            "id": 281,
            "seek": 180444,
            "start": 1813.1,
            "end": 1822.16,
            "text": " to combine the input from transformer block and another header-related input.",
            "tokens": [
                50376,
                281,
                10432,
                264,
                4846,
                490,
                31782,
                3461,
                293,
                1071,
                23117,
                12,
                12004,
                4846,
                13,
                50760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25615026595744683,
            "compression_ratio": 1.4225352112676057,
            "no_speech_prob": 0.025787353515625,
            "words": [
                {
                    "start": 1813.1,
                    "end": 1813.6,
                    "word": " to",
                    "probability": 0.0906982421875
                },
                {
                    "start": 1813.6,
                    "end": 1814.18,
                    "word": " combine",
                    "probability": 0.85693359375
                },
                {
                    "start": 1814.18,
                    "end": 1814.48,
                    "word": " the",
                    "probability": 0.7890625
                },
                {
                    "start": 1814.48,
                    "end": 1814.92,
                    "word": " input",
                    "probability": 0.75830078125
                },
                {
                    "start": 1814.92,
                    "end": 1815.38,
                    "word": " from",
                    "probability": 0.81103515625
                },
                {
                    "start": 1815.38,
                    "end": 1816.44,
                    "word": " transformer",
                    "probability": 0.51123046875
                },
                {
                    "start": 1816.44,
                    "end": 1816.96,
                    "word": " block",
                    "probability": 0.7607421875
                },
                {
                    "start": 1816.96,
                    "end": 1817.76,
                    "word": " and",
                    "probability": 0.82177734375
                },
                {
                    "start": 1817.76,
                    "end": 1818.34,
                    "word": " another",
                    "probability": 0.83349609375
                },
                {
                    "start": 1818.34,
                    "end": 1819.94,
                    "word": " header",
                    "probability": 0.069091796875
                },
                {
                    "start": 1819.94,
                    "end": 1820.46,
                    "word": "-related",
                    "probability": 0.79052734375
                },
                {
                    "start": 1820.46,
                    "end": 1821.12,
                    "word": " input",
                    "probability": 0.86962890625
                },
                {
                    "start": 1821.12,
                    "end": 1822.16,
                    "word": ".",
                    "probability": 0.908203125
                }
            ]
        },
        {
            "id": 282,
            "seek": 180444,
            "start": 1822.68,
            "end": 1828.88,
            "text": " With the additional input, we improved our model's performance further.",
            "tokens": [
                50852,
                2022,
                264,
                4497,
                4846,
                11,
                321,
                9689,
                527,
                2316,
                311,
                3389,
                3052,
                13,
                51128
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25615026595744683,
            "compression_ratio": 1.4225352112676057,
            "no_speech_prob": 0.025787353515625,
            "words": [
                {
                    "start": 1822.68,
                    "end": 1823.2,
                    "word": " With",
                    "probability": 0.87451171875
                },
                {
                    "start": 1823.2,
                    "end": 1823.32,
                    "word": " the",
                    "probability": 0.3857421875
                },
                {
                    "start": 1823.32,
                    "end": 1823.7,
                    "word": " additional",
                    "probability": 0.83544921875
                },
                {
                    "start": 1823.7,
                    "end": 1824.22,
                    "word": " input",
                    "probability": 0.86865234375
                },
                {
                    "start": 1824.22,
                    "end": 1825.14,
                    "word": ",",
                    "probability": 0.87939453125
                },
                {
                    "start": 1825.14,
                    "end": 1825.72,
                    "word": " we",
                    "probability": 0.904296875
                },
                {
                    "start": 1825.72,
                    "end": 1826.2,
                    "word": " improved",
                    "probability": 0.71044921875
                },
                {
                    "start": 1826.2,
                    "end": 1826.86,
                    "word": " our",
                    "probability": 0.8017578125
                },
                {
                    "start": 1826.86,
                    "end": 1827.36,
                    "word": " model's",
                    "probability": 0.830810546875
                },
                {
                    "start": 1827.36,
                    "end": 1827.84,
                    "word": " performance",
                    "probability": 0.82421875
                },
                {
                    "start": 1827.84,
                    "end": 1828.36,
                    "word": " further",
                    "probability": 0.86767578125
                },
                {
                    "start": 1828.36,
                    "end": 1828.88,
                    "word": ".",
                    "probability": 0.9833984375
                }
            ]
        },
        {
            "id": 283,
            "seek": 180444,
            "start": 1834.01,
            "end": 1838.75,
            "text": " Let me talk about the details of our trans adapters.",
            "tokens": [
                51212,
                961,
                385,
                751,
                466,
                264,
                4365,
                295,
                527,
                1145,
                23169,
                1559,
                13,
                51428
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25615026595744683,
            "compression_ratio": 1.4225352112676057,
            "no_speech_prob": 0.025787353515625,
            "words": [
                {
                    "start": 1834.01,
                    "end": 1834.27,
                    "word": " Let",
                    "probability": 0.861328125
                },
                {
                    "start": 1834.27,
                    "end": 1834.37,
                    "word": " me",
                    "probability": 0.89599609375
                },
                {
                    "start": 1834.37,
                    "end": 1834.67,
                    "word": " talk",
                    "probability": 0.8056640625
                },
                {
                    "start": 1834.67,
                    "end": 1835.11,
                    "word": " about",
                    "probability": 0.8115234375
                },
                {
                    "start": 1835.11,
                    "end": 1835.83,
                    "word": " the",
                    "probability": 0.79541015625
                },
                {
                    "start": 1835.83,
                    "end": 1836.21,
                    "word": " details",
                    "probability": 0.697265625
                },
                {
                    "start": 1836.21,
                    "end": 1836.57,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1836.57,
                    "end": 1836.87,
                    "word": " our",
                    "probability": 0.76953125
                },
                {
                    "start": 1836.87,
                    "end": 1837.29,
                    "word": " trans",
                    "probability": 0.322998046875
                },
                {
                    "start": 1837.29,
                    "end": 1838.23,
                    "word": " adapters",
                    "probability": 0.755126953125
                },
                {
                    "start": 1838.23,
                    "end": 1838.75,
                    "word": ".",
                    "probability": 0.95751953125
                }
            ]
        },
        {
            "id": 284,
            "seek": 182740,
            "start": 1840.35,
            "end": 1848.31,
            "text": " We inserted two adapters here and each adapter has quite a simple architecture.",
            "tokens": [
                50412,
                492,
                27992,
                732,
                23169,
                1559,
                510,
                293,
                1184,
                22860,
                575,
                1596,
                257,
                2199,
                9482,
                13,
                50732
            ],
            "temperature": 0.0,
            "avg_logprob": -0.32438150731225807,
            "compression_ratio": 1.4855072463768115,
            "no_speech_prob": 0.01441192626953125,
            "words": [
                {
                    "start": 1840.35,
                    "end": 1841.33,
                    "word": " We",
                    "probability": 0.671875
                },
                {
                    "start": 1841.33,
                    "end": 1841.93,
                    "word": " inserted",
                    "probability": 0.87646484375
                },
                {
                    "start": 1841.93,
                    "end": 1842.43,
                    "word": " two",
                    "probability": 0.7744140625
                },
                {
                    "start": 1842.43,
                    "end": 1843.13,
                    "word": " adapters",
                    "probability": 0.88134765625
                },
                {
                    "start": 1843.13,
                    "end": 1843.61,
                    "word": " here",
                    "probability": 0.767578125
                },
                {
                    "start": 1843.61,
                    "end": 1844.33,
                    "word": " and",
                    "probability": 0.35498046875
                },
                {
                    "start": 1844.33,
                    "end": 1844.65,
                    "word": " each",
                    "probability": 0.8544921875
                },
                {
                    "start": 1844.65,
                    "end": 1845.11,
                    "word": " adapter",
                    "probability": 0.75830078125
                },
                {
                    "start": 1845.11,
                    "end": 1845.47,
                    "word": " has",
                    "probability": 0.8974609375
                },
                {
                    "start": 1845.47,
                    "end": 1845.87,
                    "word": " quite",
                    "probability": 0.83154296875
                },
                {
                    "start": 1846.05,
                    "end": 1846.49,
                    "word": " a simple",
                    "probability": 0.92529296875
                },
                {
                    "start": 1846.49,
                    "end": 1847.33,
                    "word": " architecture",
                    "probability": 0.81689453125
                },
                {
                    "start": 1847.33,
                    "end": 1848.31,
                    "word": ".",
                    "probability": 0.8681640625
                }
            ]
        },
        {
            "id": 285,
            "seek": 182740,
            "start": 1849.05,
            "end": 1857.49,
            "text": " Each adapter will have two dense layers and there is one non-linear activation unit in between",
            "tokens": [
                50844,
                6947,
                22860,
                486,
                362,
                732,
                18011,
                7914,
                293,
                456,
                307,
                472,
                2107,
                12,
                28263,
                24433,
                4985,
                294,
                1296,
                51252
            ],
            "temperature": 0.0,
            "avg_logprob": -0.32438150731225807,
            "compression_ratio": 1.4855072463768115,
            "no_speech_prob": 0.01441192626953125,
            "words": [
                {
                    "start": 1849.05,
                    "end": 1849.81,
                    "word": " Each",
                    "probability": 0.84423828125
                },
                {
                    "start": 1849.81,
                    "end": 1850.33,
                    "word": " adapter",
                    "probability": 0.78515625
                },
                {
                    "start": 1850.33,
                    "end": 1851.27,
                    "word": " will",
                    "probability": 0.802734375
                },
                {
                    "start": 1851.27,
                    "end": 1851.53,
                    "word": " have",
                    "probability": 0.8828125
                },
                {
                    "start": 1851.53,
                    "end": 1851.85,
                    "word": " two",
                    "probability": 0.76416015625
                },
                {
                    "start": 1851.85,
                    "end": 1852.35,
                    "word": " dense",
                    "probability": 0.796875
                },
                {
                    "start": 1852.35,
                    "end": 1852.93,
                    "word": " layers",
                    "probability": 0.92041015625
                },
                {
                    "start": 1852.93,
                    "end": 1853.55,
                    "word": " and",
                    "probability": 0.80029296875
                },
                {
                    "start": 1853.55,
                    "end": 1853.89,
                    "word": " there",
                    "probability": 0.826171875
                },
                {
                    "start": 1853.89,
                    "end": 1854.01,
                    "word": " is",
                    "probability": 0.7109375
                },
                {
                    "start": 1854.01,
                    "end": 1854.43,
                    "word": " one",
                    "probability": 0.771484375
                },
                {
                    "start": 1854.43,
                    "end": 1855.27,
                    "word": " non",
                    "probability": 0.921875
                },
                {
                    "start": 1855.27,
                    "end": 1855.61,
                    "word": "-linear",
                    "probability": 0.7177734375
                },
                {
                    "start": 1855.61,
                    "end": 1856.31,
                    "word": " activation",
                    "probability": 0.87158203125
                },
                {
                    "start": 1856.31,
                    "end": 1856.79,
                    "word": " unit",
                    "probability": 0.84228515625
                },
                {
                    "start": 1856.79,
                    "end": 1857.11,
                    "word": " in",
                    "probability": 0.88134765625
                },
                {
                    "start": 1857.11,
                    "end": 1857.49,
                    "word": " between",
                    "probability": 0.75439453125
                }
            ]
        },
        {
            "id": 286,
            "seek": 182740,
            "start": 1857.49,
            "end": 1861.71,
            "text": " and we have a skip connection.",
            "tokens": [
                51284,
                293,
                321,
                362,
                257,
                10023,
                4984,
                13,
                51408
            ],
            "temperature": 0.0,
            "avg_logprob": -0.32438150731225807,
            "compression_ratio": 1.4855072463768115,
            "no_speech_prob": 0.01441192626953125,
            "words": [
                {
                    "start": 1857.49,
                    "end": 1858.47,
                    "word": " and",
                    "probability": 0.50341796875
                },
                {
                    "start": 1858.47,
                    "end": 1858.71,
                    "word": " we",
                    "probability": 0.88720703125
                },
                {
                    "start": 1858.71,
                    "end": 1859.07,
                    "word": " have",
                    "probability": 0.87744140625
                },
                {
                    "start": 1859.79,
                    "end": 1860.17,
                    "word": " a skip",
                    "probability": 0.767578125
                },
                {
                    "start": 1860.17,
                    "end": 1860.73,
                    "word": " connection",
                    "probability": 0.76904296875
                },
                {
                    "start": 1860.73,
                    "end": 1861.71,
                    "word": ".",
                    "probability": 0.927734375
                }
            ]
        },
        {
            "id": 287,
            "seek": 184614,
            "start": 1862.93,
            "end": 1871.39,
            "text": " The dimensionality of a dense unit is the same as the output of transformer blocks.",
            "tokens": [
                50408,
                440,
                10139,
                1860,
                295,
                257,
                18011,
                4985,
                307,
                264,
                912,
                382,
                264,
                5598,
                295,
                31782,
                8474,
                13,
                50760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.37571021711284464,
            "compression_ratio": 1.372093023255814,
            "no_speech_prob": 0.00937652587890625,
            "words": [
                {
                    "start": 1862.93,
                    "end": 1863.75,
                    "word": " The",
                    "probability": 0.67236328125
                },
                {
                    "start": 1863.75,
                    "end": 1864.57,
                    "word": " dimensionality",
                    "probability": 0.849609375
                },
                {
                    "start": 1864.57,
                    "end": 1864.81,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 1865.01,
                    "end": 1865.27,
                    "word": " a dense",
                    "probability": 0.861328125
                },
                {
                    "start": 1865.27,
                    "end": 1865.63,
                    "word": " unit",
                    "probability": 0.875
                },
                {
                    "start": 1865.63,
                    "end": 1866.61,
                    "word": " is",
                    "probability": 0.380126953125
                },
                {
                    "start": 1866.61,
                    "end": 1867.77,
                    "word": " the",
                    "probability": 0.37646484375
                },
                {
                    "start": 1867.77,
                    "end": 1867.83,
                    "word": " same",
                    "probability": 0.76611328125
                },
                {
                    "start": 1867.83,
                    "end": 1868.17,
                    "word": " as",
                    "probability": 0.87548828125
                },
                {
                    "start": 1868.17,
                    "end": 1868.59,
                    "word": " the",
                    "probability": 0.6591796875
                },
                {
                    "start": 1868.59,
                    "end": 1868.97,
                    "word": " output",
                    "probability": 0.7607421875
                },
                {
                    "start": 1868.97,
                    "end": 1869.33,
                    "word": " of",
                    "probability": 0.89111328125
                },
                {
                    "start": 1869.33,
                    "end": 1870.03,
                    "word": " transformer",
                    "probability": 0.446044921875
                },
                {
                    "start": 1870.03,
                    "end": 1870.57,
                    "word": " blocks",
                    "probability": 0.8203125
                },
                {
                    "start": 1870.57,
                    "end": 1871.39,
                    "word": ".",
                    "probability": 0.70751953125
                }
            ]
        },
        {
            "id": 288,
            "seek": 184614,
            "start": 1871.69,
            "end": 1880.45,
            "text": " And the two dense layers are initialized with near-zero values. So as initially, the adapters",
            "tokens": [
                50832,
                400,
                264,
                732,
                18011,
                7914,
                366,
                5883,
                1602,
                365,
                2651,
                12,
                32226,
                4190,
                13,
                407,
                382,
                9105,
                11,
                264,
                23169,
                1559,
                51260
            ],
            "temperature": 0.0,
            "avg_logprob": -0.37571021711284464,
            "compression_ratio": 1.372093023255814,
            "no_speech_prob": 0.00937652587890625,
            "words": [
                {
                    "start": 1871.69,
                    "end": 1872.51,
                    "word": " And",
                    "probability": 0.253173828125
                },
                {
                    "start": 1872.51,
                    "end": 1872.83,
                    "word": " the",
                    "probability": 0.6357421875
                },
                {
                    "start": 1872.83,
                    "end": 1873.19,
                    "word": " two",
                    "probability": 0.78759765625
                },
                {
                    "start": 1873.19,
                    "end": 1873.53,
                    "word": " dense",
                    "probability": 0.82177734375
                },
                {
                    "start": 1873.53,
                    "end": 1873.97,
                    "word": " layers",
                    "probability": 0.91064453125
                },
                {
                    "start": 1873.97,
                    "end": 1874.29,
                    "word": " are",
                    "probability": 0.875
                },
                {
                    "start": 1874.29,
                    "end": 1875.29,
                    "word": " initialized",
                    "probability": 0.940185546875
                },
                {
                    "start": 1875.29,
                    "end": 1875.61,
                    "word": " with",
                    "probability": 0.8046875
                },
                {
                    "start": 1875.61,
                    "end": 1876.01,
                    "word": " near",
                    "probability": 0.75390625
                },
                {
                    "start": 1876.01,
                    "end": 1876.23,
                    "word": "-zero",
                    "probability": 0.6904296875
                },
                {
                    "start": 1876.23,
                    "end": 1876.81,
                    "word": " values",
                    "probability": 0.92822265625
                },
                {
                    "start": 1876.81,
                    "end": 1877.41,
                    "word": ".",
                    "probability": 0.7841796875
                },
                {
                    "start": 1877.41,
                    "end": 1877.63,
                    "word": " So",
                    "probability": 0.6904296875
                },
                {
                    "start": 1877.63,
                    "end": 1878.41,
                    "word": " as",
                    "probability": 0.490478515625
                },
                {
                    "start": 1878.41,
                    "end": 1878.91,
                    "word": " initially",
                    "probability": 0.79345703125
                },
                {
                    "start": 1878.91,
                    "end": 1879.65,
                    "word": ",",
                    "probability": 0.65625
                },
                {
                    "start": 1879.65,
                    "end": 1879.85,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 1879.85,
                    "end": 1880.45,
                    "word": " adapters",
                    "probability": 0.8388671875
                }
            ]
        },
        {
            "id": 289,
            "seek": 187614,
            "start": 1880.45,
            "end": 1892.45,
            "text": " will act as identity blocks, but however, they will gradually change the data from the lower",
            "tokens": [
                50368,
                486,
                605,
                382,
                6575,
                8474,
                11,
                457,
                4461,
                11,
                436,
                486,
                13145,
                1319,
                264,
                1412,
                490,
                264,
                3126,
                50976
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2266995676776819,
            "compression_ratio": 1.6375,
            "no_speech_prob": 0.0516357421875,
            "words": [
                {
                    "start": 1880.45,
                    "end": 1880.81,
                    "word": " will",
                    "probability": 0.5546875
                },
                {
                    "start": 1880.81,
                    "end": 1881.25,
                    "word": " act",
                    "probability": 0.77392578125
                },
                {
                    "start": 1881.25,
                    "end": 1881.73,
                    "word": " as",
                    "probability": 0.89794921875
                },
                {
                    "start": 1881.73,
                    "end": 1884.21,
                    "word": " identity",
                    "probability": 0.7412109375
                },
                {
                    "start": 1884.21,
                    "end": 1885.17,
                    "word": " blocks",
                    "probability": 0.8330078125
                },
                {
                    "start": 1885.17,
                    "end": 1885.57,
                    "word": ",",
                    "probability": 0.41259765625
                },
                {
                    "start": 1885.57,
                    "end": 1885.65,
                    "word": " but",
                    "probability": 0.7373046875
                },
                {
                    "start": 1885.65,
                    "end": 1886.13,
                    "word": " however",
                    "probability": 0.7333984375
                },
                {
                    "start": 1886.13,
                    "end": 1887.01,
                    "word": ",",
                    "probability": 0.5419921875
                },
                {
                    "start": 1887.01,
                    "end": 1887.27,
                    "word": " they",
                    "probability": 0.73291015625
                },
                {
                    "start": 1887.27,
                    "end": 1887.51,
                    "word": " will",
                    "probability": 0.81689453125
                },
                {
                    "start": 1887.51,
                    "end": 1888.13,
                    "word": " gradually",
                    "probability": 0.8359375
                },
                {
                    "start": 1888.13,
                    "end": 1889.15,
                    "word": " change",
                    "probability": 0.84130859375
                },
                {
                    "start": 1889.15,
                    "end": 1889.93,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 1889.93,
                    "end": 1890.79,
                    "word": " data",
                    "probability": 0.78515625
                },
                {
                    "start": 1890.79,
                    "end": 1891.85,
                    "word": " from",
                    "probability": 0.8134765625
                },
                {
                    "start": 1891.85,
                    "end": 1892.13,
                    "word": " the",
                    "probability": 0.7890625
                },
                {
                    "start": 1892.13,
                    "end": 1892.45,
                    "word": " lower",
                    "probability": 0.630859375
                }
            ]
        },
        {
            "id": 290,
            "seek": 187614,
            "start": 1892.45,
            "end": 1901.13,
            "text": " transformer block to upper transformer block and to minimize classification loss.",
            "tokens": [
                50976,
                31782,
                3461,
                281,
                6597,
                31782,
                3461,
                293,
                281,
                17522,
                21538,
                4470,
                13,
                51344
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2266995676776819,
            "compression_ratio": 1.6375,
            "no_speech_prob": 0.0516357421875,
            "words": [
                {
                    "start": 1892.45,
                    "end": 1893.31,
                    "word": " transformer",
                    "probability": 0.40380859375
                },
                {
                    "start": 1893.31,
                    "end": 1893.77,
                    "word": " block",
                    "probability": 0.77685546875
                },
                {
                    "start": 1893.77,
                    "end": 1894.15,
                    "word": " to",
                    "probability": 0.88818359375
                },
                {
                    "start": 1894.15,
                    "end": 1894.59,
                    "word": " upper",
                    "probability": 0.66455078125
                },
                {
                    "start": 1894.59,
                    "end": 1895.15,
                    "word": " transformer",
                    "probability": 0.75146484375
                },
                {
                    "start": 1895.15,
                    "end": 1895.63,
                    "word": " block",
                    "probability": 0.76904296875
                },
                {
                    "start": 1895.63,
                    "end": 1896.59,
                    "word": " and",
                    "probability": 0.43603515625
                },
                {
                    "start": 1896.59,
                    "end": 1897.01,
                    "word": " to",
                    "probability": 0.82568359375
                },
                {
                    "start": 1897.01,
                    "end": 1898.57,
                    "word": " minimize",
                    "probability": 0.83251953125
                },
                {
                    "start": 1898.57,
                    "end": 1899.39,
                    "word": " classification",
                    "probability": 0.91796875
                },
                {
                    "start": 1899.39,
                    "end": 1900.17,
                    "word": " loss",
                    "probability": 0.7978515625
                },
                {
                    "start": 1900.17,
                    "end": 1901.13,
                    "word": ".",
                    "probability": 0.97216796875
                }
            ]
        },
        {
            "id": 291,
            "seek": 187614,
            "start": 1905.58,
            "end": 1913.38,
            "text": " We also modified the standard fine-tuning method by using a partial fine-tuning method.",
            "tokens": [
                51436,
                492,
                611,
                15873,
                264,
                3832,
                2489,
                12,
                83,
                37726,
                3170,
                538,
                1228,
                257,
                14641,
                2489,
                12,
                83,
                37726,
                3170,
                13,
                51816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2266995676776819,
            "compression_ratio": 1.6375,
            "no_speech_prob": 0.0516357421875,
            "words": [
                {
                    "start": 1905.58,
                    "end": 1905.88,
                    "word": " We",
                    "probability": 0.89697265625
                },
                {
                    "start": 1905.88,
                    "end": 1906.4,
                    "word": " also",
                    "probability": 0.7998046875
                },
                {
                    "start": 1906.4,
                    "end": 1907.04,
                    "word": " modified",
                    "probability": 0.650390625
                },
                {
                    "start": 1907.04,
                    "end": 1907.76,
                    "word": " the",
                    "probability": 0.64111328125
                },
                {
                    "start": 1907.76,
                    "end": 1908.18,
                    "word": " standard",
                    "probability": 0.8671875
                },
                {
                    "start": 1908.18,
                    "end": 1908.62,
                    "word": " fine",
                    "probability": 0.77734375
                },
                {
                    "start": 1908.62,
                    "end": 1908.92,
                    "word": "-tuning",
                    "probability": 0.8225911458333334
                },
                {
                    "start": 1908.92,
                    "end": 1909.6,
                    "word": " method",
                    "probability": 0.8359375
                },
                {
                    "start": 1909.6,
                    "end": 1910.0,
                    "word": " by",
                    "probability": 0.9052734375
                },
                {
                    "start": 1910.0,
                    "end": 1910.6,
                    "word": " using",
                    "probability": 0.83056640625
                },
                {
                    "start": 1911.58,
                    "end": 1911.96,
                    "word": " a partial",
                    "probability": 0.876953125
                },
                {
                    "start": 1911.96,
                    "end": 1912.32,
                    "word": " fine",
                    "probability": 0.818359375
                },
                {
                    "start": 1912.32,
                    "end": 1912.62,
                    "word": "-tuning",
                    "probability": 0.92529296875
                },
                {
                    "start": 1912.62,
                    "end": 1913.24,
                    "word": " method",
                    "probability": 0.86181640625
                },
                {
                    "start": 1913.24,
                    "end": 1913.38,
                    "word": ".",
                    "probability": 0.908203125
                }
            ]
        },
        {
            "id": 292,
            "seek": 190312,
            "start": 1914.8,
            "end": 1922.74,
            "text": " Standard fine-tuning involves updating all parameters jointly. However, for partial",
            "tokens": [
                50408,
                21298,
                2489,
                12,
                83,
                37726,
                11626,
                25113,
                439,
                9834,
                46557,
                13,
                2908,
                11,
                337,
                14641,
                50792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24278846153846154,
            "compression_ratio": 1.5481927710843373,
            "no_speech_prob": 0.004604339599609375,
            "words": [
                {
                    "start": 1914.8,
                    "end": 1915.64,
                    "word": " Standard",
                    "probability": 0.830078125
                },
                {
                    "start": 1915.64,
                    "end": 1916.08,
                    "word": " fine",
                    "probability": 0.74658203125
                },
                {
                    "start": 1916.08,
                    "end": 1916.38,
                    "word": "-tuning",
                    "probability": 0.8177083333333334
                },
                {
                    "start": 1916.38,
                    "end": 1917.08,
                    "word": " involves",
                    "probability": 0.8427734375
                },
                {
                    "start": 1917.08,
                    "end": 1917.82,
                    "word": " updating",
                    "probability": 0.68359375
                },
                {
                    "start": 1917.82,
                    "end": 1918.36,
                    "word": " all",
                    "probability": 0.9013671875
                },
                {
                    "start": 1918.36,
                    "end": 1919.16,
                    "word": " parameters",
                    "probability": 0.8564453125
                },
                {
                    "start": 1919.16,
                    "end": 1920.1,
                    "word": " jointly",
                    "probability": 0.84619140625
                },
                {
                    "start": 1920.1,
                    "end": 1920.94,
                    "word": ".",
                    "probability": 0.798828125
                },
                {
                    "start": 1921.14,
                    "end": 1921.76,
                    "word": " However",
                    "probability": 0.85888671875
                },
                {
                    "start": 1921.76,
                    "end": 1922.32,
                    "word": ",",
                    "probability": 0.9248046875
                },
                {
                    "start": 1922.32,
                    "end": 1922.5,
                    "word": " for",
                    "probability": 0.6845703125
                },
                {
                    "start": 1922.5,
                    "end": 1922.74,
                    "word": " partial",
                    "probability": 0.00576019287109375
                }
            ]
        },
        {
            "id": 293,
            "seek": 190312,
            "start": 1922.74,
            "end": 1931.7,
            "text": " fine-tuning, we only update upper blocks, but we fix low blocks. For example, here,",
            "tokens": [
                50792,
                2489,
                12,
                83,
                37726,
                11,
                321,
                787,
                5623,
                6597,
                8474,
                11,
                457,
                321,
                3191,
                2295,
                8474,
                13,
                1171,
                1365,
                11,
                510,
                11,
                51192
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24278846153846154,
            "compression_ratio": 1.5481927710843373,
            "no_speech_prob": 0.004604339599609375,
            "words": [
                {
                    "start": 1922.74,
                    "end": 1923.04,
                    "word": " fine",
                    "probability": 0.86474609375
                },
                {
                    "start": 1923.04,
                    "end": 1923.3,
                    "word": "-tuning",
                    "probability": 0.9195963541666666
                },
                {
                    "start": 1923.3,
                    "end": 1923.9,
                    "word": ",",
                    "probability": 0.888671875
                },
                {
                    "start": 1923.9,
                    "end": 1924.14,
                    "word": " we",
                    "probability": 0.9033203125
                },
                {
                    "start": 1924.14,
                    "end": 1925.04,
                    "word": " only",
                    "probability": 0.779296875
                },
                {
                    "start": 1925.04,
                    "end": 1925.58,
                    "word": " update",
                    "probability": 0.75048828125
                },
                {
                    "start": 1925.58,
                    "end": 1926.38,
                    "word": " upper",
                    "probability": 0.77490234375
                },
                {
                    "start": 1926.38,
                    "end": 1926.96,
                    "word": " blocks",
                    "probability": 0.845703125
                },
                {
                    "start": 1926.96,
                    "end": 1927.3,
                    "word": ",",
                    "probability": 0.6630859375
                },
                {
                    "start": 1927.3,
                    "end": 1927.52,
                    "word": " but",
                    "probability": 0.8876953125
                },
                {
                    "start": 1927.52,
                    "end": 1927.9,
                    "word": " we",
                    "probability": 0.82666015625
                },
                {
                    "start": 1927.9,
                    "end": 1928.16,
                    "word": " fix",
                    "probability": 0.880859375
                },
                {
                    "start": 1928.16,
                    "end": 1928.5,
                    "word": " low",
                    "probability": 0.443603515625
                },
                {
                    "start": 1928.5,
                    "end": 1928.92,
                    "word": " blocks",
                    "probability": 0.74658203125
                },
                {
                    "start": 1928.92,
                    "end": 1929.7,
                    "word": ".",
                    "probability": 0.96337890625
                },
                {
                    "start": 1929.7,
                    "end": 1929.94,
                    "word": " For",
                    "probability": 0.89111328125
                },
                {
                    "start": 1929.94,
                    "end": 1930.3,
                    "word": " example",
                    "probability": 0.93017578125
                },
                {
                    "start": 1930.3,
                    "end": 1930.48,
                    "word": ",",
                    "probability": 0.615234375
                },
                {
                    "start": 1930.48,
                    "end": 1930.74,
                    "word": " here",
                    "probability": 0.77197265625
                },
                {
                    "start": 1930.74,
                    "end": 1931.7,
                    "word": ",",
                    "probability": 0.60107421875
                }
            ]
        },
        {
            "id": 294,
            "seek": 190312,
            "start": 1931.7,
            "end": 1941.1,
            "text": " the low blocks embedding the transformer 1 and 3 are fixed, but we update adapter 1 and 2",
            "tokens": [
                51240,
                264,
                2295,
                8474,
                12240,
                3584,
                264,
                31782,
                502,
                293,
                805,
                366,
                6806,
                11,
                457,
                321,
                5623,
                22860,
                502,
                293,
                568,
                51712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24278846153846154,
            "compression_ratio": 1.5481927710843373,
            "no_speech_prob": 0.004604339599609375,
            "words": [
                {
                    "start": 1931.7,
                    "end": 1931.96,
                    "word": " the",
                    "probability": 0.79541015625
                },
                {
                    "start": 1931.96,
                    "end": 1932.34,
                    "word": " low",
                    "probability": 0.74609375
                },
                {
                    "start": 1932.34,
                    "end": 1932.76,
                    "word": " blocks",
                    "probability": 0.76708984375
                },
                {
                    "start": 1932.76,
                    "end": 1933.72,
                    "word": " embedding",
                    "probability": 0.84423828125
                },
                {
                    "start": 1933.72,
                    "end": 1933.98,
                    "word": " the",
                    "probability": 0.79833984375
                },
                {
                    "start": 1933.98,
                    "end": 1934.64,
                    "word": " transformer",
                    "probability": 0.54541015625
                },
                {
                    "start": 1934.64,
                    "end": 1935.04,
                    "word": " 1",
                    "probability": 0.876953125
                },
                {
                    "start": 1935.04,
                    "end": 1935.22,
                    "word": " and",
                    "probability": 0.74365234375
                },
                {
                    "start": 1935.22,
                    "end": 1935.76,
                    "word": " 3",
                    "probability": 0.98974609375
                },
                {
                    "start": 1935.76,
                    "end": 1937.18,
                    "word": " are",
                    "probability": 0.85546875
                },
                {
                    "start": 1937.18,
                    "end": 1937.5,
                    "word": " fixed",
                    "probability": 0.822265625
                },
                {
                    "start": 1937.5,
                    "end": 1937.86,
                    "word": ",",
                    "probability": 0.830078125
                },
                {
                    "start": 1937.86,
                    "end": 1938.36,
                    "word": " but",
                    "probability": 0.90234375
                },
                {
                    "start": 1938.36,
                    "end": 1938.8,
                    "word": " we",
                    "probability": 0.90283203125
                },
                {
                    "start": 1938.8,
                    "end": 1939.5,
                    "word": " update",
                    "probability": 0.68212890625
                },
                {
                    "start": 1939.5,
                    "end": 1940.32,
                    "word": " adapter",
                    "probability": 0.60107421875
                },
                {
                    "start": 1940.32,
                    "end": 1940.6,
                    "word": " 1",
                    "probability": 0.96533203125
                },
                {
                    "start": 1940.6,
                    "end": 1940.76,
                    "word": " and",
                    "probability": 0.87939453125
                },
                {
                    "start": 1940.76,
                    "end": 1941.1,
                    "word": " 2",
                    "probability": 0.994140625
                }
            ]
        },
        {
            "id": 295,
            "seek": 192799,
            "start": 1941.1,
            "end": 1951.24,
            "text": " and transformer 5 and classification header. This approach was to minimize forgetting",
            "tokens": [
                50364,
                293,
                31782,
                1025,
                293,
                21538,
                23117,
                13,
                639,
                3109,
                390,
                281,
                17522,
                25428,
                50784
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2582908163265306,
            "compression_ratio": 1.4484848484848485,
            "no_speech_prob": 0.007030487060546875,
            "words": [
                {
                    "start": 1941.1,
                    "end": 1941.34,
                    "word": " and",
                    "probability": 0.1463623046875
                },
                {
                    "start": 1941.34,
                    "end": 1941.84,
                    "word": " transformer",
                    "probability": 0.005344390869140625
                },
                {
                    "start": 1941.84,
                    "end": 1942.46,
                    "word": " 5",
                    "probability": 0.61865234375
                },
                {
                    "start": 1942.46,
                    "end": 1943.04,
                    "word": " and",
                    "probability": 0.6552734375
                },
                {
                    "start": 1943.04,
                    "end": 1943.92,
                    "word": " classification",
                    "probability": 0.87548828125
                },
                {
                    "start": 1943.92,
                    "end": 1944.46,
                    "word": " header",
                    "probability": 0.69482421875
                },
                {
                    "start": 1944.46,
                    "end": 1945.48,
                    "word": ".",
                    "probability": 0.93017578125
                },
                {
                    "start": 1948.24,
                    "end": 1948.74,
                    "word": " This",
                    "probability": 0.79736328125
                },
                {
                    "start": 1948.74,
                    "end": 1949.18,
                    "word": " approach",
                    "probability": 0.89990234375
                },
                {
                    "start": 1949.18,
                    "end": 1949.48,
                    "word": " was",
                    "probability": 0.8916015625
                },
                {
                    "start": 1949.48,
                    "end": 1949.82,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 1949.82,
                    "end": 1950.26,
                    "word": " minimize",
                    "probability": 0.728515625
                },
                {
                    "start": 1950.26,
                    "end": 1951.24,
                    "word": " forgetting",
                    "probability": 0.7646484375
                }
            ]
        },
        {
            "id": 296,
            "seek": 192799,
            "start": 1951.24,
            "end": 1959.4,
            "text": " problems of learned presentations from low transformer blocks.",
            "tokens": [
                50784,
                2740,
                295,
                3264,
                18964,
                490,
                2295,
                31782,
                8474,
                13,
                51120
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2582908163265306,
            "compression_ratio": 1.4484848484848485,
            "no_speech_prob": 0.007030487060546875,
            "words": [
                {
                    "start": 1951.24,
                    "end": 1952.36,
                    "word": " problems",
                    "probability": 0.70556640625
                },
                {
                    "start": 1952.36,
                    "end": 1952.88,
                    "word": " of",
                    "probability": 0.87060546875
                },
                {
                    "start": 1952.88,
                    "end": 1953.76,
                    "word": " learned",
                    "probability": 0.79150390625
                },
                {
                    "start": 1953.76,
                    "end": 1954.64,
                    "word": " presentations",
                    "probability": 0.7626953125
                },
                {
                    "start": 1954.64,
                    "end": 1956.66,
                    "word": " from",
                    "probability": 0.7744140625
                },
                {
                    "start": 1956.66,
                    "end": 1957.4,
                    "word": " low",
                    "probability": 0.438232421875
                },
                {
                    "start": 1957.4,
                    "end": 1957.92,
                    "word": " transformer",
                    "probability": 0.79150390625
                },
                {
                    "start": 1957.92,
                    "end": 1958.38,
                    "word": " blocks",
                    "probability": 0.82958984375
                },
                {
                    "start": 1958.38,
                    "end": 1959.4,
                    "word": ".",
                    "probability": 0.95654296875
                }
            ]
        },
        {
            "id": 297,
            "seek": 192799,
            "start": 1962.23,
            "end": 1970.43,
            "text": " As mentioned earlier, we have two set of features. One from text data and another one from",
            "tokens": [
                51208,
                1018,
                2835,
                3071,
                11,
                321,
                362,
                732,
                992,
                295,
                4122,
                13,
                1485,
                490,
                2487,
                1412,
                293,
                1071,
                472,
                490,
                51596
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2582908163265306,
            "compression_ratio": 1.4484848484848485,
            "no_speech_prob": 0.007030487060546875,
            "words": [
                {
                    "start": 1962.23,
                    "end": 1962.74,
                    "word": " As",
                    "probability": 0.89208984375
                },
                {
                    "start": 1962.74,
                    "end": 1963.1,
                    "word": " mentioned",
                    "probability": 0.75341796875
                },
                {
                    "start": 1963.1,
                    "end": 1963.52,
                    "word": " earlier",
                    "probability": 0.90283203125
                },
                {
                    "start": 1963.52,
                    "end": 1963.52,
                    "word": ",",
                    "probability": 0.76318359375
                },
                {
                    "start": 1963.52,
                    "end": 1963.84,
                    "word": " we",
                    "probability": 0.908203125
                },
                {
                    "start": 1963.84,
                    "end": 1964.18,
                    "word": " have",
                    "probability": 0.8720703125
                },
                {
                    "start": 1964.18,
                    "end": 1964.44,
                    "word": " two",
                    "probability": 0.76953125
                },
                {
                    "start": 1964.44,
                    "end": 1964.76,
                    "word": " set",
                    "probability": 0.7109375
                },
                {
                    "start": 1964.76,
                    "end": 1964.96,
                    "word": " of",
                    "probability": 0.89208984375
                },
                {
                    "start": 1964.96,
                    "end": 1966.12,
                    "word": " features",
                    "probability": 0.67626953125
                },
                {
                    "start": 1966.12,
                    "end": 1966.72,
                    "word": ".",
                    "probability": 0.353515625
                },
                {
                    "start": 1966.72,
                    "end": 1967.24,
                    "word": " One",
                    "probability": 0.8876953125
                },
                {
                    "start": 1967.24,
                    "end": 1968.0,
                    "word": " from",
                    "probability": 0.79541015625
                },
                {
                    "start": 1968.0,
                    "end": 1968.44,
                    "word": " text",
                    "probability": 0.85546875
                },
                {
                    "start": 1968.44,
                    "end": 1968.82,
                    "word": " data",
                    "probability": 0.8330078125
                },
                {
                    "start": 1968.82,
                    "end": 1969.24,
                    "word": " and",
                    "probability": 0.7880859375
                },
                {
                    "start": 1969.24,
                    "end": 1969.64,
                    "word": " another",
                    "probability": 0.8095703125
                },
                {
                    "start": 1969.64,
                    "end": 1969.92,
                    "word": " one",
                    "probability": 0.806640625
                },
                {
                    "start": 1969.92,
                    "end": 1970.43,
                    "word": " from",
                    "probability": 0.82177734375
                }
            ]
        },
        {
            "id": 298,
            "seek": 194923,
            "start": 1971.29,
            "end": 1980.41,
            "text": " email headers. We can use multiple email header builders, for example, from two cc-reply",
            "tokens": [
                50412,
                3796,
                45101,
                13,
                492,
                393,
                764,
                3866,
                3796,
                23117,
                36281,
                11,
                337,
                1365,
                11,
                490,
                732,
                269,
                66,
                12,
                265,
                2724,
                50856
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3154069670411043,
            "compression_ratio": 1.4375,
            "no_speech_prob": 0.00506591796875,
            "words": [
                {
                    "start": 1971.29,
                    "end": 1971.85,
                    "word": " email",
                    "probability": 0.109130859375
                },
                {
                    "start": 1971.85,
                    "end": 1972.27,
                    "word": " headers",
                    "probability": 0.86865234375
                },
                {
                    "start": 1972.27,
                    "end": 1972.93,
                    "word": ".",
                    "probability": 0.89599609375
                },
                {
                    "start": 1972.93,
                    "end": 1973.25,
                    "word": " We",
                    "probability": 0.86083984375
                },
                {
                    "start": 1973.25,
                    "end": 1973.47,
                    "word": " can",
                    "probability": 0.8837890625
                },
                {
                    "start": 1973.47,
                    "end": 1973.83,
                    "word": " use",
                    "probability": 0.826171875
                },
                {
                    "start": 1973.83,
                    "end": 1975.71,
                    "word": " multiple",
                    "probability": 0.72802734375
                },
                {
                    "start": 1975.71,
                    "end": 1976.31,
                    "word": " email",
                    "probability": 0.919921875
                },
                {
                    "start": 1976.31,
                    "end": 1976.89,
                    "word": " header",
                    "probability": 0.6689453125
                },
                {
                    "start": 1976.89,
                    "end": 1977.27,
                    "word": " builders",
                    "probability": 0.49169921875
                },
                {
                    "start": 1977.27,
                    "end": 1977.43,
                    "word": ",",
                    "probability": 0.62353515625
                },
                {
                    "start": 1977.43,
                    "end": 1977.61,
                    "word": " for",
                    "probability": 0.88037109375
                },
                {
                    "start": 1977.61,
                    "end": 1977.97,
                    "word": " example",
                    "probability": 0.9248046875
                },
                {
                    "start": 1977.97,
                    "end": 1978.25,
                    "word": ",",
                    "probability": 0.67724609375
                },
                {
                    "start": 1978.25,
                    "end": 1978.55,
                    "word": " from",
                    "probability": 0.43896484375
                },
                {
                    "start": 1978.55,
                    "end": 1979.23,
                    "word": " two",
                    "probability": 0.7666015625
                },
                {
                    "start": 1979.23,
                    "end": 1979.85,
                    "word": " cc",
                    "probability": 0.667724609375
                },
                {
                    "start": 1979.85,
                    "end": 1980.41,
                    "word": "-reply",
                    "probability": 0.6363932291666666
                }
            ]
        },
        {
            "id": 299,
            "seek": 194923,
            "start": 1981.37,
            "end": 1991.67,
            "text": " builders to extract additional context information. And we consider the subject and the text as",
            "tokens": [
                50916,
                36281,
                281,
                8947,
                4497,
                4319,
                1589,
                13,
                400,
                321,
                1949,
                264,
                3983,
                293,
                264,
                2487,
                382,
                51416
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3154069670411043,
            "compression_ratio": 1.4375,
            "no_speech_prob": 0.00506591796875,
            "words": [
                {
                    "start": 1981.37,
                    "end": 1981.93,
                    "word": " builders",
                    "probability": 0.44091796875
                },
                {
                    "start": 1981.93,
                    "end": 1982.51,
                    "word": " to",
                    "probability": 0.84228515625
                },
                {
                    "start": 1982.51,
                    "end": 1984.17,
                    "word": " extract",
                    "probability": 0.734375
                },
                {
                    "start": 1984.17,
                    "end": 1984.85,
                    "word": " additional",
                    "probability": 0.84375
                },
                {
                    "start": 1984.85,
                    "end": 1985.41,
                    "word": " context",
                    "probability": 0.74755859375
                },
                {
                    "start": 1985.41,
                    "end": 1986.19,
                    "word": " information",
                    "probability": 0.71533203125
                },
                {
                    "start": 1986.19,
                    "end": 1987.27,
                    "word": ".",
                    "probability": 0.46826171875
                },
                {
                    "start": 1987.27,
                    "end": 1987.39,
                    "word": " And",
                    "probability": 0.61767578125
                },
                {
                    "start": 1987.39,
                    "end": 1987.57,
                    "word": " we",
                    "probability": 0.85693359375
                },
                {
                    "start": 1987.57,
                    "end": 1988.09,
                    "word": " consider",
                    "probability": 0.58056640625
                },
                {
                    "start": 1988.09,
                    "end": 1988.53,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 1988.53,
                    "end": 1988.95,
                    "word": " subject",
                    "probability": 0.8466796875
                },
                {
                    "start": 1988.95,
                    "end": 1989.57,
                    "word": " and",
                    "probability": 0.86328125
                },
                {
                    "start": 1989.57,
                    "end": 1990.51,
                    "word": " the",
                    "probability": 0.748046875
                },
                {
                    "start": 1990.51,
                    "end": 1991.13,
                    "word": " text",
                    "probability": 0.87646484375
                },
                {
                    "start": 1991.13,
                    "end": 1991.67,
                    "word": " as",
                    "probability": 0.88623046875
                }
            ]
        },
        {
            "id": 300,
            "seek": 197849,
            "start": 1992.17,
            "end": 2005.03,
            "text": " context input. The first set of features are from email text content features.",
            "tokens": [
                50392,
                4319,
                4846,
                13,
                440,
                700,
                992,
                295,
                4122,
                366,
                490,
                3796,
                2487,
                2701,
                4122,
                13,
                50848
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19921874683914761,
            "compression_ratio": 1.6114649681528663,
            "no_speech_prob": 0.030364990234375,
            "words": [
                {
                    "start": 1992.17,
                    "end": 1992.65,
                    "word": " context",
                    "probability": 0.04034423828125
                },
                {
                    "start": 1992.65,
                    "end": 1993.21,
                    "word": " input",
                    "probability": 0.75732421875
                },
                {
                    "start": 1993.21,
                    "end": 1994.17,
                    "word": ".",
                    "probability": 0.78271484375
                },
                {
                    "start": 1997.45,
                    "end": 1997.87,
                    "word": " The",
                    "probability": 0.6796875
                },
                {
                    "start": 1997.87,
                    "end": 1998.23,
                    "word": " first",
                    "probability": 0.79150390625
                },
                {
                    "start": 1998.23,
                    "end": 1999.83,
                    "word": " set",
                    "probability": 0.8291015625
                },
                {
                    "start": 1999.83,
                    "end": 1999.97,
                    "word": " of",
                    "probability": 0.90966796875
                },
                {
                    "start": 1999.97,
                    "end": 2000.23,
                    "word": " features",
                    "probability": 0.61376953125
                },
                {
                    "start": 2000.23,
                    "end": 2000.55,
                    "word": " are",
                    "probability": 0.734375
                },
                {
                    "start": 2000.55,
                    "end": 2000.81,
                    "word": " from",
                    "probability": 0.77001953125
                },
                {
                    "start": 2000.81,
                    "end": 2001.51,
                    "word": " email",
                    "probability": 0.77685546875
                },
                {
                    "start": 2001.51,
                    "end": 2002.05,
                    "word": " text",
                    "probability": 0.82177734375
                },
                {
                    "start": 2002.05,
                    "end": 2003.37,
                    "word": " content",
                    "probability": 0.439697265625
                },
                {
                    "start": 2003.37,
                    "end": 2004.07,
                    "word": " features",
                    "probability": 0.6962890625
                },
                {
                    "start": 2004.07,
                    "end": 2005.03,
                    "word": ".",
                    "probability": 0.88720703125
                }
            ]
        },
        {
            "id": 301,
            "seek": 197849,
            "start": 2005.27,
            "end": 2015.05,
            "text": " We extract text data from subject and plain text body. If only html content is available,",
            "tokens": [
                50912,
                492,
                8947,
                2487,
                1412,
                490,
                3983,
                293,
                11121,
                2487,
                1772,
                13,
                759,
                787,
                276,
                83,
                15480,
                2701,
                307,
                2435,
                11,
                51388
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19921874683914761,
            "compression_ratio": 1.6114649681528663,
            "no_speech_prob": 0.030364990234375,
            "words": [
                {
                    "start": 2005.27,
                    "end": 2005.67,
                    "word": " We",
                    "probability": 0.857421875
                },
                {
                    "start": 2005.67,
                    "end": 2006.27,
                    "word": " extract",
                    "probability": 0.72216796875
                },
                {
                    "start": 2006.27,
                    "end": 2007.01,
                    "word": " text",
                    "probability": 0.90625
                },
                {
                    "start": 2007.01,
                    "end": 2007.35,
                    "word": " data",
                    "probability": 0.8203125
                },
                {
                    "start": 2007.35,
                    "end": 2008.17,
                    "word": " from",
                    "probability": 0.8173828125
                },
                {
                    "start": 2008.17,
                    "end": 2008.71,
                    "word": " subject",
                    "probability": 0.68310546875
                },
                {
                    "start": 2008.71,
                    "end": 2009.05,
                    "word": " and",
                    "probability": 0.8896484375
                },
                {
                    "start": 2009.05,
                    "end": 2009.41,
                    "word": " plain",
                    "probability": 0.87939453125
                },
                {
                    "start": 2009.41,
                    "end": 2009.79,
                    "word": " text",
                    "probability": 0.8828125
                },
                {
                    "start": 2009.79,
                    "end": 2010.19,
                    "word": " body",
                    "probability": 0.73974609375
                },
                {
                    "start": 2010.19,
                    "end": 2011.15,
                    "word": ".",
                    "probability": 0.91552734375
                },
                {
                    "start": 2011.75,
                    "end": 2012.15,
                    "word": " If",
                    "probability": 0.88232421875
                },
                {
                    "start": 2012.15,
                    "end": 2012.79,
                    "word": " only",
                    "probability": 0.83544921875
                },
                {
                    "start": 2012.79,
                    "end": 2013.35,
                    "word": " html",
                    "probability": 0.6793009440104166
                },
                {
                    "start": 2013.35,
                    "end": 2013.83,
                    "word": " content",
                    "probability": 0.8125
                },
                {
                    "start": 2013.83,
                    "end": 2014.17,
                    "word": " is",
                    "probability": 0.90625
                },
                {
                    "start": 2014.17,
                    "end": 2014.75,
                    "word": " available",
                    "probability": 0.75
                },
                {
                    "start": 2014.75,
                    "end": 2015.05,
                    "word": ",",
                    "probability": 0.685546875
                }
            ]
        },
        {
            "id": 302,
            "seek": 197849,
            "start": 2015.05,
            "end": 2023.65,
            "text": " then we can also extract plain text from html data using a html parser. For example,",
            "tokens": [
                51388,
                550,
                321,
                393,
                611,
                8947,
                11121,
                2487,
                490,
                276,
                83,
                15480,
                1412,
                1228,
                257,
                276,
                83,
                15480,
                21156,
                260,
                13,
                1171,
                1365,
                11,
                51804
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19921874683914761,
            "compression_ratio": 1.6114649681528663,
            "no_speech_prob": 0.030364990234375,
            "words": [
                {
                    "start": 2015.05,
                    "end": 2015.27,
                    "word": " then",
                    "probability": 0.6953125
                },
                {
                    "start": 2015.27,
                    "end": 2015.69,
                    "word": " we",
                    "probability": 0.880859375
                },
                {
                    "start": 2015.69,
                    "end": 2015.91,
                    "word": " can",
                    "probability": 0.88671875
                },
                {
                    "start": 2015.91,
                    "end": 2016.21,
                    "word": " also",
                    "probability": 0.81982421875
                },
                {
                    "start": 2016.21,
                    "end": 2016.71,
                    "word": " extract",
                    "probability": 0.74658203125
                },
                {
                    "start": 2016.71,
                    "end": 2017.13,
                    "word": " plain",
                    "probability": 0.89501953125
                },
                {
                    "start": 2017.13,
                    "end": 2017.49,
                    "word": " text",
                    "probability": 0.89453125
                },
                {
                    "start": 2017.49,
                    "end": 2018.03,
                    "word": " from",
                    "probability": 0.80419921875
                },
                {
                    "start": 2018.03,
                    "end": 2018.93,
                    "word": " html",
                    "probability": 0.93017578125
                },
                {
                    "start": 2018.93,
                    "end": 2019.91,
                    "word": " data",
                    "probability": 0.80029296875
                },
                {
                    "start": 2019.91,
                    "end": 2020.49,
                    "word": " using",
                    "probability": 0.7998046875
                },
                {
                    "start": 2020.73,
                    "end": 2021.25,
                    "word": " a html",
                    "probability": 0.9358723958333334
                },
                {
                    "start": 2021.25,
                    "end": 2021.73,
                    "word": " parser",
                    "probability": 0.888671875
                },
                {
                    "start": 2021.73,
                    "end": 2022.31,
                    "word": ".",
                    "probability": 0.9521484375
                },
                {
                    "start": 2022.31,
                    "end": 2022.79,
                    "word": " For",
                    "probability": 0.89013671875
                },
                {
                    "start": 2022.79,
                    "end": 2023.17,
                    "word": " example",
                    "probability": 0.93017578125
                },
                {
                    "start": 2023.17,
                    "end": 2023.65,
                    "word": ",",
                    "probability": 0.892578125
                }
            ]
        },
        {
            "id": 303,
            "seek": 200561,
            "start": 2023.65,
            "end": 2037.13,
            "text": " this one is a simple html hyperlink, but we only extract visible text visit bank site as output.",
            "tokens": [
                50380,
                341,
                472,
                307,
                257,
                2199,
                276,
                83,
                15480,
                9848,
                22473,
                11,
                457,
                321,
                787,
                8947,
                8974,
                2487,
                3441,
                3765,
                3621,
                382,
                5598,
                13,
                50992
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23808262711864406,
            "compression_ratio": 1.503030303030303,
            "no_speech_prob": 0.001033782958984375,
            "words": [
                {
                    "start": 2023.65,
                    "end": 2024.25,
                    "word": " this",
                    "probability": 0.03131103515625
                },
                {
                    "start": 2024.25,
                    "end": 2024.49,
                    "word": " one",
                    "probability": 0.69140625
                },
                {
                    "start": 2024.49,
                    "end": 2024.73,
                    "word": " is",
                    "probability": 0.876953125
                },
                {
                    "start": 2025.55,
                    "end": 2026.25,
                    "word": " a simple",
                    "probability": 0.72998046875
                },
                {
                    "start": 2026.25,
                    "end": 2027.79,
                    "word": " html",
                    "probability": 0.7445475260416666
                },
                {
                    "start": 2027.79,
                    "end": 2028.59,
                    "word": " hyperlink",
                    "probability": 0.9072265625
                },
                {
                    "start": 2028.59,
                    "end": 2029.65,
                    "word": ",",
                    "probability": 0.255126953125
                },
                {
                    "start": 2029.65,
                    "end": 2029.91,
                    "word": " but",
                    "probability": 0.79833984375
                },
                {
                    "start": 2029.91,
                    "end": 2030.19,
                    "word": " we",
                    "probability": 0.8857421875
                },
                {
                    "start": 2030.19,
                    "end": 2030.55,
                    "word": " only",
                    "probability": 0.81298828125
                },
                {
                    "start": 2030.55,
                    "end": 2030.95,
                    "word": " extract",
                    "probability": 0.666015625
                },
                {
                    "start": 2030.95,
                    "end": 2031.59,
                    "word": " visible",
                    "probability": 0.6640625
                },
                {
                    "start": 2031.59,
                    "end": 2033.21,
                    "word": " text",
                    "probability": 0.81201171875
                },
                {
                    "start": 2033.21,
                    "end": 2034.15,
                    "word": " visit",
                    "probability": 0.107666015625
                },
                {
                    "start": 2034.15,
                    "end": 2034.57,
                    "word": " bank",
                    "probability": 0.6240234375
                },
                {
                    "start": 2034.57,
                    "end": 2035.03,
                    "word": " site",
                    "probability": 0.77734375
                },
                {
                    "start": 2035.03,
                    "end": 2035.73,
                    "word": " as",
                    "probability": 0.857421875
                },
                {
                    "start": 2035.73,
                    "end": 2036.21,
                    "word": " output",
                    "probability": 0.763671875
                },
                {
                    "start": 2036.21,
                    "end": 2037.13,
                    "word": ".",
                    "probability": 0.6064453125
                }
            ]
        },
        {
            "id": 304,
            "seek": 200561,
            "start": 2037.91,
            "end": 2044.15,
            "text": " And then from the extracted text, we remove less informative characters.",
            "tokens": [
                51080,
                400,
                550,
                490,
                264,
                34086,
                2487,
                11,
                321,
                4159,
                1570,
                27759,
                4342,
                13,
                51368
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23808262711864406,
            "compression_ratio": 1.503030303030303,
            "no_speech_prob": 0.001033782958984375,
            "words": [
                {
                    "start": 2037.91,
                    "end": 2038.27,
                    "word": " And",
                    "probability": 0.59619140625
                },
                {
                    "start": 2038.27,
                    "end": 2038.49,
                    "word": " then",
                    "probability": 0.71533203125
                },
                {
                    "start": 2038.49,
                    "end": 2038.83,
                    "word": " from",
                    "probability": 0.7109375
                },
                {
                    "start": 2038.83,
                    "end": 2039.01,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 2039.01,
                    "end": 2039.45,
                    "word": " extracted",
                    "probability": 0.533203125
                },
                {
                    "start": 2039.45,
                    "end": 2039.85,
                    "word": " text",
                    "probability": 0.88037109375
                },
                {
                    "start": 2039.85,
                    "end": 2040.03,
                    "word": ",",
                    "probability": 0.474609375
                },
                {
                    "start": 2040.03,
                    "end": 2040.13,
                    "word": " we",
                    "probability": 0.90087890625
                },
                {
                    "start": 2040.13,
                    "end": 2040.65,
                    "word": " remove",
                    "probability": 0.62841796875
                },
                {
                    "start": 2040.65,
                    "end": 2042.23,
                    "word": " less",
                    "probability": 0.76904296875
                },
                {
                    "start": 2042.23,
                    "end": 2042.83,
                    "word": " informative",
                    "probability": 0.95263671875
                },
                {
                    "start": 2042.83,
                    "end": 2043.79,
                    "word": " characters",
                    "probability": 0.7568359375
                },
                {
                    "start": 2043.79,
                    "end": 2044.15,
                    "word": ".",
                    "probability": 0.76904296875
                }
            ]
        },
        {
            "id": 305,
            "seek": 200561,
            "start": 2044.87,
            "end": 2050.77,
            "text": " We remove digit and punctuation characters. The second example, there are many",
            "tokens": [
                51436,
                492,
                4159,
                14293,
                293,
                27006,
                16073,
                4342,
                13,
                440,
                1150,
                1365,
                11,
                456,
                366,
                867,
                51736
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23808262711864406,
            "compression_ratio": 1.503030303030303,
            "no_speech_prob": 0.001033782958984375,
            "words": [
                {
                    "start": 2044.87,
                    "end": 2045.33,
                    "word": " We",
                    "probability": 0.85693359375
                },
                {
                    "start": 2045.33,
                    "end": 2045.79,
                    "word": " remove",
                    "probability": 0.84716796875
                },
                {
                    "start": 2045.79,
                    "end": 2046.93,
                    "word": " digit",
                    "probability": 0.76123046875
                },
                {
                    "start": 2046.93,
                    "end": 2047.39,
                    "word": " and",
                    "probability": 0.892578125
                },
                {
                    "start": 2047.39,
                    "end": 2047.99,
                    "word": " punctuation",
                    "probability": 0.769287109375
                },
                {
                    "start": 2047.99,
                    "end": 2048.63,
                    "word": " characters",
                    "probability": 0.74755859375
                },
                {
                    "start": 2048.63,
                    "end": 2048.89,
                    "word": ".",
                    "probability": 0.93359375
                },
                {
                    "start": 2048.89,
                    "end": 2049.05,
                    "word": " The",
                    "probability": 0.73974609375
                },
                {
                    "start": 2049.05,
                    "end": 2049.33,
                    "word": " second",
                    "probability": 0.79638671875
                },
                {
                    "start": 2049.33,
                    "end": 2049.87,
                    "word": " example",
                    "probability": 0.94091796875
                },
                {
                    "start": 2049.87,
                    "end": 2050.07,
                    "word": ",",
                    "probability": 0.802734375
                },
                {
                    "start": 2050.07,
                    "end": 2050.23,
                    "word": " there",
                    "probability": 0.8486328125
                },
                {
                    "start": 2050.23,
                    "end": 2050.37,
                    "word": " are",
                    "probability": 0.89111328125
                },
                {
                    "start": 2050.37,
                    "end": 2050.77,
                    "word": " many",
                    "probability": 0.8154296875
                }
            ]
        },
        {
            "id": 306,
            "seek": 202807,
            "start": 2050.77,
            "end": 2061.03,
            "text": " hyphens, but the hyphens will be removed in this step. Otherwise, each hyphen will be",
            "tokens": [
                50380,
                2477,
                950,
                694,
                11,
                457,
                264,
                2477,
                950,
                694,
                486,
                312,
                7261,
                294,
                341,
                1823,
                13,
                10328,
                11,
                1184,
                2477,
                47059,
                486,
                312,
                50868
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21093749735090467,
            "compression_ratio": 1.330708661417323,
            "no_speech_prob": 0.038818359375,
            "words": [
                {
                    "start": 2050.77,
                    "end": 2051.77,
                    "word": " hyphens",
                    "probability": 0.635430653889974
                },
                {
                    "start": 2051.77,
                    "end": 2051.91,
                    "word": ",",
                    "probability": 0.10699462890625
                },
                {
                    "start": 2051.91,
                    "end": 2052.19,
                    "word": " but",
                    "probability": 0.67919921875
                },
                {
                    "start": 2052.19,
                    "end": 2052.79,
                    "word": " the",
                    "probability": 0.63623046875
                },
                {
                    "start": 2052.79,
                    "end": 2053.29,
                    "word": " hyphens",
                    "probability": 0.9031575520833334
                },
                {
                    "start": 2053.29,
                    "end": 2053.63,
                    "word": " will",
                    "probability": 0.80029296875
                },
                {
                    "start": 2053.63,
                    "end": 2053.81,
                    "word": " be",
                    "probability": 0.9033203125
                },
                {
                    "start": 2053.81,
                    "end": 2054.23,
                    "word": " removed",
                    "probability": 0.771484375
                },
                {
                    "start": 2054.23,
                    "end": 2054.57,
                    "word": " in",
                    "probability": 0.84423828125
                },
                {
                    "start": 2054.57,
                    "end": 2055.83,
                    "word": " this",
                    "probability": 0.810546875
                },
                {
                    "start": 2055.83,
                    "end": 2056.23,
                    "word": " step",
                    "probability": 0.86279296875
                },
                {
                    "start": 2056.23,
                    "end": 2057.23,
                    "word": ".",
                    "probability": 0.85888671875
                },
                {
                    "start": 2057.23,
                    "end": 2057.59,
                    "word": " Otherwise",
                    "probability": 0.77490234375
                },
                {
                    "start": 2057.59,
                    "end": 2057.93,
                    "word": ",",
                    "probability": 0.75927734375
                },
                {
                    "start": 2057.93,
                    "end": 2058.21,
                    "word": " each",
                    "probability": 0.83544921875
                },
                {
                    "start": 2058.21,
                    "end": 2058.73,
                    "word": " hyphen",
                    "probability": 0.91015625
                },
                {
                    "start": 2058.73,
                    "end": 2059.67,
                    "word": " will",
                    "probability": 0.79736328125
                },
                {
                    "start": 2059.67,
                    "end": 2061.03,
                    "word": " be",
                    "probability": 0.87158203125
                }
            ]
        },
        {
            "id": 307,
            "seek": 202807,
            "start": 2061.81,
            "end": 2073.23,
            "text": " individual token. Finally, we select 120 tokens as input to the transformer blocks.",
            "tokens": [
                50916,
                2609,
                14862,
                13,
                6288,
                11,
                321,
                3048,
                10411,
                22667,
                382,
                4846,
                281,
                264,
                31782,
                8474,
                13,
                51436
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21093749735090467,
            "compression_ratio": 1.330708661417323,
            "no_speech_prob": 0.038818359375,
            "words": [
                {
                    "start": 2061.81,
                    "end": 2062.55,
                    "word": " individual",
                    "probability": 0.1253662109375
                },
                {
                    "start": 2062.55,
                    "end": 2063.29,
                    "word": " token",
                    "probability": 0.67138671875
                },
                {
                    "start": 2063.29,
                    "end": 2064.41,
                    "word": ".",
                    "probability": 0.91650390625
                },
                {
                    "start": 2064.85,
                    "end": 2065.97,
                    "word": " Finally",
                    "probability": 0.6875
                },
                {
                    "start": 2065.97,
                    "end": 2066.11,
                    "word": ",",
                    "probability": 0.88232421875
                },
                {
                    "start": 2066.11,
                    "end": 2066.27,
                    "word": " we",
                    "probability": 0.8955078125
                },
                {
                    "start": 2066.27,
                    "end": 2066.91,
                    "word": " select",
                    "probability": 0.75390625
                },
                {
                    "start": 2066.91,
                    "end": 2067.99,
                    "word": " 120",
                    "probability": 0.86279296875
                },
                {
                    "start": 2067.99,
                    "end": 2068.75,
                    "word": " tokens",
                    "probability": 0.7373046875
                },
                {
                    "start": 2068.75,
                    "end": 2069.63,
                    "word": " as",
                    "probability": 0.869140625
                },
                {
                    "start": 2069.63,
                    "end": 2070.05,
                    "word": " input",
                    "probability": 0.8916015625
                },
                {
                    "start": 2070.05,
                    "end": 2070.37,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 2070.37,
                    "end": 2070.63,
                    "word": " the",
                    "probability": 0.8017578125
                },
                {
                    "start": 2070.63,
                    "end": 2071.27,
                    "word": " transformer",
                    "probability": 0.744140625
                },
                {
                    "start": 2071.27,
                    "end": 2072.31,
                    "word": " blocks",
                    "probability": 0.69384765625
                },
                {
                    "start": 2072.31,
                    "end": 2073.23,
                    "word": ".",
                    "probability": 0.9619140625
                }
            ]
        },
        {
            "id": 308,
            "seek": 205807,
            "start": 2077.29,
            "end": 2084.87,
            "text": " We use a sub-word tokenized called WordPiece. The WordPiece tokenized can overcome",
            "tokens": [
                50408,
                492,
                764,
                257,
                1422,
                12,
                7462,
                14862,
                1602,
                1219,
                8725,
                47,
                46566,
                13,
                440,
                8725,
                47,
                46566,
                14862,
                1602,
                393,
                10473,
                50792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2845293209876543,
            "compression_ratio": 1.6476683937823835,
            "no_speech_prob": 0.04290771484375,
            "words": [
                {
                    "start": 2077.29,
                    "end": 2077.77,
                    "word": " We",
                    "probability": 0.80712890625
                },
                {
                    "start": 2077.77,
                    "end": 2078.15,
                    "word": " use",
                    "probability": 0.7490234375
                },
                {
                    "start": 2078.51,
                    "end": 2079.55,
                    "word": " a sub",
                    "probability": 0.69921875
                },
                {
                    "start": 2079.55,
                    "end": 2079.79,
                    "word": "-word",
                    "probability": 0.663330078125
                },
                {
                    "start": 2079.79,
                    "end": 2080.43,
                    "word": " tokenized",
                    "probability": 0.6356201171875
                },
                {
                    "start": 2080.43,
                    "end": 2080.89,
                    "word": " called",
                    "probability": 0.71435546875
                },
                {
                    "start": 2080.89,
                    "end": 2081.77,
                    "word": " WordPiece",
                    "probability": 0.5597330729166666
                },
                {
                    "start": 2081.77,
                    "end": 2082.39,
                    "word": ".",
                    "probability": 0.904296875
                },
                {
                    "start": 2082.39,
                    "end": 2082.73,
                    "word": " The",
                    "probability": 0.7451171875
                },
                {
                    "start": 2082.73,
                    "end": 2083.29,
                    "word": " WordPiece",
                    "probability": 0.93017578125
                },
                {
                    "start": 2083.29,
                    "end": 2083.97,
                    "word": " tokenized",
                    "probability": 0.841064453125
                },
                {
                    "start": 2083.97,
                    "end": 2084.33,
                    "word": " can",
                    "probability": 0.8955078125
                },
                {
                    "start": 2084.33,
                    "end": 2084.87,
                    "word": " overcome",
                    "probability": 0.83935546875
                }
            ]
        },
        {
            "id": 309,
            "seek": 205807,
            "start": 2084.87,
            "end": 2093.95,
            "text": " some of the limitations from character-level or word-level tokens. Character-level tokens are",
            "tokens": [
                50792,
                512,
                295,
                264,
                15705,
                490,
                2517,
                12,
                12418,
                420,
                1349,
                12,
                12418,
                22667,
                13,
                36786,
                12,
                12418,
                22667,
                366,
                51232
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2845293209876543,
            "compression_ratio": 1.6476683937823835,
            "no_speech_prob": 0.04290771484375,
            "words": [
                {
                    "start": 2084.87,
                    "end": 2085.51,
                    "word": " some",
                    "probability": 0.8037109375
                },
                {
                    "start": 2085.51,
                    "end": 2085.65,
                    "word": " of",
                    "probability": 0.83203125
                },
                {
                    "start": 2085.65,
                    "end": 2085.77,
                    "word": " the",
                    "probability": 0.755859375
                },
                {
                    "start": 2085.77,
                    "end": 2086.37,
                    "word": " limitations",
                    "probability": 0.65673828125
                },
                {
                    "start": 2086.37,
                    "end": 2087.07,
                    "word": " from",
                    "probability": 0.7802734375
                },
                {
                    "start": 2087.07,
                    "end": 2087.61,
                    "word": " character",
                    "probability": 0.53662109375
                },
                {
                    "start": 2087.61,
                    "end": 2087.99,
                    "word": "-level",
                    "probability": 0.635009765625
                },
                {
                    "start": 2087.99,
                    "end": 2088.55,
                    "word": " or",
                    "probability": 0.78466796875
                },
                {
                    "start": 2088.55,
                    "end": 2089.41,
                    "word": " word",
                    "probability": 0.77734375
                },
                {
                    "start": 2089.41,
                    "end": 2089.77,
                    "word": "-level",
                    "probability": 0.93212890625
                },
                {
                    "start": 2089.77,
                    "end": 2090.69,
                    "word": " tokens",
                    "probability": 0.71484375
                },
                {
                    "start": 2090.69,
                    "end": 2091.65,
                    "word": ".",
                    "probability": 0.94287109375
                },
                {
                    "start": 2091.91,
                    "end": 2092.87,
                    "word": " Character",
                    "probability": 0.779296875
                },
                {
                    "start": 2092.87,
                    "end": 2093.23,
                    "word": "-level",
                    "probability": 0.914306640625
                },
                {
                    "start": 2093.23,
                    "end": 2093.67,
                    "word": " tokens",
                    "probability": 0.76904296875
                },
                {
                    "start": 2093.67,
                    "end": 2093.95,
                    "word": " are",
                    "probability": 0.89208984375
                }
            ]
        },
        {
            "id": 310,
            "seek": 205807,
            "start": 2093.95,
            "end": 2100.19,
            "text": " too fine-grained, so it is hard to recognize word boundaries and meaning of words.",
            "tokens": [
                51232,
                886,
                2489,
                12,
                20735,
                2001,
                11,
                370,
                309,
                307,
                1152,
                281,
                5521,
                1349,
                13180,
                293,
                3620,
                295,
                2283,
                13,
                51496
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2845293209876543,
            "compression_ratio": 1.6476683937823835,
            "no_speech_prob": 0.04290771484375,
            "words": [
                {
                    "start": 2093.95,
                    "end": 2094.29,
                    "word": " too",
                    "probability": 0.79248046875
                },
                {
                    "start": 2094.29,
                    "end": 2094.61,
                    "word": " fine",
                    "probability": 0.8505859375
                },
                {
                    "start": 2094.61,
                    "end": 2095.01,
                    "word": "-grained",
                    "probability": 0.8216145833333334
                },
                {
                    "start": 2095.01,
                    "end": 2095.31,
                    "word": ",",
                    "probability": 0.4482421875
                },
                {
                    "start": 2095.31,
                    "end": 2095.45,
                    "word": " so",
                    "probability": 0.890625
                },
                {
                    "start": 2095.45,
                    "end": 2095.73,
                    "word": " it",
                    "probability": 0.8955078125
                },
                {
                    "start": 2095.73,
                    "end": 2095.89,
                    "word": " is",
                    "probability": 0.8447265625
                },
                {
                    "start": 2095.89,
                    "end": 2096.19,
                    "word": " hard",
                    "probability": 0.8232421875
                },
                {
                    "start": 2096.19,
                    "end": 2096.49,
                    "word": " to",
                    "probability": 0.90869140625
                },
                {
                    "start": 2096.49,
                    "end": 2097.03,
                    "word": " recognize",
                    "probability": 0.8154296875
                },
                {
                    "start": 2097.03,
                    "end": 2097.41,
                    "word": " word",
                    "probability": 0.79296875
                },
                {
                    "start": 2097.41,
                    "end": 2097.95,
                    "word": " boundaries",
                    "probability": 0.8359375
                },
                {
                    "start": 2097.95,
                    "end": 2098.31,
                    "word": " and",
                    "probability": 0.8837890625
                },
                {
                    "start": 2098.31,
                    "end": 2098.63,
                    "word": " meaning",
                    "probability": 0.56396484375
                },
                {
                    "start": 2098.63,
                    "end": 2098.81,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 2098.81,
                    "end": 2099.29,
                    "word": " words",
                    "probability": 0.59521484375
                },
                {
                    "start": 2099.29,
                    "end": 2100.19,
                    "word": ".",
                    "probability": 0.63330078125
                }
            ]
        },
        {
            "id": 311,
            "seek": 205807,
            "start": 2100.19,
            "end": 2106.37,
            "text": " And word-level tokens often have auto-vocabulary problems.",
            "tokens": [
                51536,
                400,
                1349,
                12,
                12418,
                22667,
                2049,
                362,
                8399,
                12,
                20836,
                19189,
                2740,
                13,
                51824
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2845293209876543,
            "compression_ratio": 1.6476683937823835,
            "no_speech_prob": 0.04290771484375,
            "words": [
                {
                    "start": 2100.19,
                    "end": 2100.51,
                    "word": " And",
                    "probability": 0.67333984375
                },
                {
                    "start": 2100.51,
                    "end": 2101.39,
                    "word": " word",
                    "probability": 0.6376953125
                },
                {
                    "start": 2101.39,
                    "end": 2101.75,
                    "word": "-level",
                    "probability": 0.94482421875
                },
                {
                    "start": 2101.75,
                    "end": 2102.31,
                    "word": " tokens",
                    "probability": 0.74365234375
                },
                {
                    "start": 2102.31,
                    "end": 2103.07,
                    "word": " often",
                    "probability": 0.73583984375
                },
                {
                    "start": 2103.07,
                    "end": 2103.75,
                    "word": " have",
                    "probability": 0.8896484375
                },
                {
                    "start": 2103.75,
                    "end": 2104.69,
                    "word": " auto",
                    "probability": 0.6796875
                },
                {
                    "start": 2104.69,
                    "end": 2105.17,
                    "word": "-vocabulary",
                    "probability": 0.8326822916666666
                },
                {
                    "start": 2105.17,
                    "end": 2105.87,
                    "word": " problems",
                    "probability": 0.70068359375
                },
                {
                    "start": 2105.87,
                    "end": 2106.37,
                    "word": ".",
                    "probability": 0.98974609375
                }
            ]
        },
        {
            "id": 312,
            "seek": 207855,
            "start": 2107.32,
            "end": 2116.09,
            "text": " The sub-word token tokenized can split complex or uncommon words into sub-tokens.",
            "tokens": [
                50408,
                440,
                1422,
                12,
                7462,
                14862,
                14862,
                1602,
                393,
                7472,
                3997,
                420,
                29289,
                2283,
                666,
                1422,
                12,
                83,
                453,
                694,
                13,
                50780
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2842793270033233,
            "compression_ratio": 1.323076923076923,
            "no_speech_prob": 0.005695343017578125,
            "words": [
                {
                    "start": 2107.32,
                    "end": 2107.87,
                    "word": " The",
                    "probability": 0.65087890625
                },
                {
                    "start": 2107.87,
                    "end": 2108.13,
                    "word": " sub",
                    "probability": 0.72998046875
                },
                {
                    "start": 2108.13,
                    "end": 2108.37,
                    "word": "-word",
                    "probability": 0.587158203125
                },
                {
                    "start": 2108.37,
                    "end": 2108.95,
                    "word": " token",
                    "probability": 0.81201171875
                },
                {
                    "start": 2108.95,
                    "end": 2109.97,
                    "word": " tokenized",
                    "probability": 0.3016357421875
                },
                {
                    "start": 2109.97,
                    "end": 2110.41,
                    "word": " can",
                    "probability": 0.87841796875
                },
                {
                    "start": 2110.41,
                    "end": 2111.05,
                    "word": " split",
                    "probability": 0.80615234375
                },
                {
                    "start": 2111.05,
                    "end": 2112.41,
                    "word": " complex",
                    "probability": 0.73046875
                },
                {
                    "start": 2112.41,
                    "end": 2112.69,
                    "word": " or",
                    "probability": 0.8232421875
                },
                {
                    "start": 2112.69,
                    "end": 2113.23,
                    "word": " uncommon",
                    "probability": 0.7509765625
                },
                {
                    "start": 2113.23,
                    "end": 2113.63,
                    "word": " words",
                    "probability": 0.71044921875
                },
                {
                    "start": 2113.63,
                    "end": 2114.01,
                    "word": " into",
                    "probability": 0.748046875
                },
                {
                    "start": 2114.01,
                    "end": 2114.39,
                    "word": " sub",
                    "probability": 0.63818359375
                },
                {
                    "start": 2114.39,
                    "end": 2114.99,
                    "word": "-tokens",
                    "probability": 0.9158935546875
                },
                {
                    "start": 2114.99,
                    "end": 2116.09,
                    "word": ".",
                    "probability": 0.873046875
                }
            ]
        },
        {
            "id": 313,
            "seek": 207855,
            "start": 2116.47,
            "end": 2127.09,
            "text": " For example, CAPBOT can be divided into simple CAP and BOT, and double HES is an indicator",
            "tokens": [
                50852,
                1171,
                1365,
                11,
                33636,
                33,
                5068,
                393,
                312,
                6666,
                666,
                2199,
                33636,
                293,
                363,
                5068,
                11,
                293,
                3834,
                389,
                2358,
                307,
                364,
                16961,
                51404
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2842793270033233,
            "compression_ratio": 1.323076923076923,
            "no_speech_prob": 0.005695343017578125,
            "words": [
                {
                    "start": 2116.47,
                    "end": 2116.67,
                    "word": " For",
                    "probability": 0.7998046875
                },
                {
                    "start": 2116.67,
                    "end": 2117.23,
                    "word": " example",
                    "probability": 0.91943359375
                },
                {
                    "start": 2117.23,
                    "end": 2117.59,
                    "word": ",",
                    "probability": 0.8779296875
                },
                {
                    "start": 2117.59,
                    "end": 2118.27,
                    "word": " CAPBOT",
                    "probability": 0.5675048828125
                },
                {
                    "start": 2118.27,
                    "end": 2119.25,
                    "word": " can",
                    "probability": 0.86376953125
                },
                {
                    "start": 2119.25,
                    "end": 2119.43,
                    "word": " be",
                    "probability": 0.90087890625
                },
                {
                    "start": 2119.43,
                    "end": 2119.85,
                    "word": " divided",
                    "probability": 0.74658203125
                },
                {
                    "start": 2119.85,
                    "end": 2120.41,
                    "word": " into",
                    "probability": 0.72802734375
                },
                {
                    "start": 2120.41,
                    "end": 2121.07,
                    "word": " simple",
                    "probability": 0.6630859375
                },
                {
                    "start": 2121.07,
                    "end": 2121.57,
                    "word": " CAP",
                    "probability": 0.841796875
                },
                {
                    "start": 2121.57,
                    "end": 2122.09,
                    "word": " and",
                    "probability": 0.8583984375
                },
                {
                    "start": 2122.09,
                    "end": 2122.47,
                    "word": " BOT",
                    "probability": 0.933837890625
                },
                {
                    "start": 2122.47,
                    "end": 2123.43,
                    "word": ",",
                    "probability": 0.50732421875
                },
                {
                    "start": 2123.43,
                    "end": 2123.55,
                    "word": " and",
                    "probability": 0.86328125
                },
                {
                    "start": 2123.55,
                    "end": 2124.69,
                    "word": " double",
                    "probability": 0.603515625
                },
                {
                    "start": 2124.69,
                    "end": 2125.35,
                    "word": " HES",
                    "probability": 0.45452880859375
                },
                {
                    "start": 2125.35,
                    "end": 2125.93,
                    "word": " is",
                    "probability": 0.81787109375
                },
                {
                    "start": 2125.93,
                    "end": 2126.47,
                    "word": " an",
                    "probability": 0.6435546875
                },
                {
                    "start": 2126.47,
                    "end": 2127.09,
                    "word": " indicator",
                    "probability": 0.68359375
                }
            ]
        },
        {
            "id": 314,
            "seek": 210751,
            "start": 2127.09,
            "end": 2139.33,
            "text": " for sub-words. Similarly, SOPHOS can be divided into three tokens, SO, PH, and OS.",
            "tokens": [
                50380,
                337,
                1422,
                12,
                13832,
                13,
                13157,
                11,
                10621,
                21904,
                4367,
                393,
                312,
                6666,
                666,
                1045,
                22667,
                11,
                10621,
                11,
                16530,
                11,
                293,
                12731,
                13,
                50924
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3105769230769231,
            "compression_ratio": 1.4506172839506173,
            "no_speech_prob": 0.319091796875,
            "words": [
                {
                    "start": 2127.09,
                    "end": 2127.79,
                    "word": " for",
                    "probability": 0.061492919921875
                },
                {
                    "start": 2127.79,
                    "end": 2128.39,
                    "word": " sub",
                    "probability": 0.59619140625
                },
                {
                    "start": 2128.39,
                    "end": 2128.85,
                    "word": "-words",
                    "probability": 0.19482421875
                },
                {
                    "start": 2128.85,
                    "end": 2129.77,
                    "word": ".",
                    "probability": 0.880859375
                },
                {
                    "start": 2130.65,
                    "end": 2131.57,
                    "word": " Similarly",
                    "probability": 0.74755859375
                },
                {
                    "start": 2131.57,
                    "end": 2131.79,
                    "word": ",",
                    "probability": 0.72216796875
                },
                {
                    "start": 2131.79,
                    "end": 2132.23,
                    "word": " SOPHOS",
                    "probability": 0.5158284505208334
                },
                {
                    "start": 2132.23,
                    "end": 2132.51,
                    "word": " can",
                    "probability": 0.896484375
                },
                {
                    "start": 2132.51,
                    "end": 2132.69,
                    "word": " be",
                    "probability": 0.90576171875
                },
                {
                    "start": 2132.69,
                    "end": 2133.07,
                    "word": " divided",
                    "probability": 0.7275390625
                },
                {
                    "start": 2133.07,
                    "end": 2133.57,
                    "word": " into",
                    "probability": 0.7392578125
                },
                {
                    "start": 2133.57,
                    "end": 2134.15,
                    "word": " three",
                    "probability": 0.65478515625
                },
                {
                    "start": 2134.15,
                    "end": 2134.67,
                    "word": " tokens",
                    "probability": 0.5908203125
                },
                {
                    "start": 2134.67,
                    "end": 2135.77,
                    "word": ",",
                    "probability": 0.352294921875
                },
                {
                    "start": 2135.77,
                    "end": 2136.23,
                    "word": " SO",
                    "probability": 0.798828125
                },
                {
                    "start": 2136.23,
                    "end": 2136.49,
                    "word": ",",
                    "probability": 0.82470703125
                },
                {
                    "start": 2136.49,
                    "end": 2137.03,
                    "word": " PH",
                    "probability": 0.841796875
                },
                {
                    "start": 2137.03,
                    "end": 2137.77,
                    "word": ",",
                    "probability": 0.76806640625
                },
                {
                    "start": 2137.77,
                    "end": 2137.99,
                    "word": " and",
                    "probability": 0.853515625
                },
                {
                    "start": 2137.99,
                    "end": 2138.41,
                    "word": " OS",
                    "probability": 0.90869140625
                },
                {
                    "start": 2138.41,
                    "end": 2139.33,
                    "word": ".",
                    "probability": 0.98388671875
                }
            ]
        },
        {
            "id": 315,
            "seek": 210751,
            "start": 2140.63,
            "end": 2147.47,
            "text": " The sub-word tokenized reduce number of unknown tokens in our email data.",
            "tokens": [
                51040,
                440,
                1422,
                12,
                7462,
                14862,
                1602,
                5407,
                1230,
                295,
                9841,
                22667,
                294,
                527,
                3796,
                1412,
                13,
                51360
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3105769230769231,
            "compression_ratio": 1.4506172839506173,
            "no_speech_prob": 0.319091796875,
            "words": [
                {
                    "start": 2140.63,
                    "end": 2140.95,
                    "word": " The",
                    "probability": 0.72607421875
                },
                {
                    "start": 2140.95,
                    "end": 2141.23,
                    "word": " sub",
                    "probability": 0.80029296875
                },
                {
                    "start": 2141.23,
                    "end": 2141.45,
                    "word": "-word",
                    "probability": 0.853271484375
                },
                {
                    "start": 2141.45,
                    "end": 2142.25,
                    "word": " tokenized",
                    "probability": 0.6156005859375
                },
                {
                    "start": 2142.25,
                    "end": 2143.23,
                    "word": " reduce",
                    "probability": 0.29296875
                },
                {
                    "start": 2143.23,
                    "end": 2143.65,
                    "word": " number",
                    "probability": 0.64501953125
                },
                {
                    "start": 2143.65,
                    "end": 2144.45,
                    "word": " of",
                    "probability": 0.89599609375
                },
                {
                    "start": 2144.45,
                    "end": 2144.97,
                    "word": " unknown",
                    "probability": 0.6494140625
                },
                {
                    "start": 2144.97,
                    "end": 2145.61,
                    "word": " tokens",
                    "probability": 0.7724609375
                },
                {
                    "start": 2145.61,
                    "end": 2146.11,
                    "word": " in",
                    "probability": 0.89794921875
                },
                {
                    "start": 2146.11,
                    "end": 2146.35,
                    "word": " our",
                    "probability": 0.80517578125
                },
                {
                    "start": 2146.35,
                    "end": 2146.63,
                    "word": " email",
                    "probability": 0.884765625
                },
                {
                    "start": 2146.63,
                    "end": 2147.01,
                    "word": " data",
                    "probability": 0.81103515625
                },
                {
                    "start": 2147.01,
                    "end": 2147.47,
                    "word": ".",
                    "probability": 0.98095703125
                }
            ]
        },
        {
            "id": 316,
            "seek": 210751,
            "start": 2151.83,
            "end": 2159.15,
            "text": " With the selected tokens, the tokens will be the input to the embedding layer,",
            "tokens": [
                51444,
                2022,
                264,
                8209,
                22667,
                11,
                264,
                22667,
                486,
                312,
                264,
                4846,
                281,
                264,
                12240,
                3584,
                4583,
                11,
                51808
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3105769230769231,
            "compression_ratio": 1.4506172839506173,
            "no_speech_prob": 0.319091796875,
            "words": [
                {
                    "start": 2151.83,
                    "end": 2152.11,
                    "word": " With",
                    "probability": 0.875
                },
                {
                    "start": 2152.11,
                    "end": 2152.29,
                    "word": " the",
                    "probability": 0.75341796875
                },
                {
                    "start": 2152.29,
                    "end": 2152.67,
                    "word": " selected",
                    "probability": 0.79443359375
                },
                {
                    "start": 2152.67,
                    "end": 2153.43,
                    "word": " tokens",
                    "probability": 0.685546875
                },
                {
                    "start": 2153.43,
                    "end": 2154.21,
                    "word": ",",
                    "probability": 0.94140625
                },
                {
                    "start": 2154.21,
                    "end": 2154.77,
                    "word": " the",
                    "probability": 0.78125
                },
                {
                    "start": 2154.77,
                    "end": 2155.99,
                    "word": " tokens",
                    "probability": 0.734375
                },
                {
                    "start": 2155.99,
                    "end": 2156.49,
                    "word": " will",
                    "probability": 0.82470703125
                },
                {
                    "start": 2156.49,
                    "end": 2156.65,
                    "word": " be",
                    "probability": 0.87060546875
                },
                {
                    "start": 2156.65,
                    "end": 2156.81,
                    "word": " the",
                    "probability": 0.24853515625
                },
                {
                    "start": 2156.81,
                    "end": 2157.11,
                    "word": " input",
                    "probability": 0.9111328125
                },
                {
                    "start": 2157.11,
                    "end": 2157.29,
                    "word": " to",
                    "probability": 0.888671875
                },
                {
                    "start": 2157.29,
                    "end": 2157.47,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 2157.47,
                    "end": 2158.23,
                    "word": " embedding",
                    "probability": 0.892578125
                },
                {
                    "start": 2158.23,
                    "end": 2158.77,
                    "word": " layer",
                    "probability": 0.7685546875
                },
                {
                    "start": 2158.77,
                    "end": 2159.15,
                    "word": ",",
                    "probability": 0.37060546875
                }
            ]
        },
        {
            "id": 317,
            "seek": 213263,
            "start": 2159.15,
            "end": 2170.45,
            "text": " and then we have three transformer blocks. And each transformer block has 12 multi-head",
            "tokens": [
                50364,
                293,
                550,
                321,
                362,
                1045,
                31782,
                8474,
                13,
                400,
                1184,
                31782,
                3461,
                575,
                2272,
                4825,
                12,
                1934,
                50928
            ],
            "temperature": 0.0,
            "avg_logprob": -0.333610381217713,
            "compression_ratio": 1.6482758620689655,
            "no_speech_prob": 0.00409698486328125,
            "words": [
                {
                    "start": 2159.15,
                    "end": 2159.37,
                    "word": " and",
                    "probability": 0.24462890625
                },
                {
                    "start": 2159.37,
                    "end": 2159.63,
                    "word": " then",
                    "probability": 0.492919921875
                },
                {
                    "start": 2159.63,
                    "end": 2160.51,
                    "word": " we",
                    "probability": 0.8525390625
                },
                {
                    "start": 2160.51,
                    "end": 2160.83,
                    "word": " have",
                    "probability": 0.8779296875
                },
                {
                    "start": 2160.83,
                    "end": 2161.37,
                    "word": " three",
                    "probability": 0.6943359375
                },
                {
                    "start": 2161.37,
                    "end": 2162.87,
                    "word": " transformer",
                    "probability": 0.70361328125
                },
                {
                    "start": 2162.87,
                    "end": 2164.85,
                    "word": " blocks",
                    "probability": 0.87158203125
                },
                {
                    "start": 2164.85,
                    "end": 2165.27,
                    "word": ".",
                    "probability": 0.263671875
                },
                {
                    "start": 2165.27,
                    "end": 2165.45,
                    "word": " And",
                    "probability": 0.6552734375
                },
                {
                    "start": 2165.45,
                    "end": 2165.67,
                    "word": " each",
                    "probability": 0.83837890625
                },
                {
                    "start": 2165.67,
                    "end": 2166.37,
                    "word": " transformer",
                    "probability": 0.7939453125
                },
                {
                    "start": 2166.37,
                    "end": 2167.27,
                    "word": " block",
                    "probability": 0.77880859375
                },
                {
                    "start": 2167.27,
                    "end": 2168.81,
                    "word": " has",
                    "probability": 0.8857421875
                },
                {
                    "start": 2168.81,
                    "end": 2169.85,
                    "word": " 12",
                    "probability": 0.8095703125
                },
                {
                    "start": 2169.85,
                    "end": 2170.21,
                    "word": " multi",
                    "probability": 0.61328125
                },
                {
                    "start": 2170.21,
                    "end": 2170.45,
                    "word": "-head",
                    "probability": 0.743408203125
                }
            ]
        },
        {
            "id": 318,
            "seek": 213263,
            "start": 2170.45,
            "end": 2179.77,
            "text": " attention layers. And each individual attention layer runs contextual relationship between tokens.",
            "tokens": [
                50928,
                3202,
                7914,
                13,
                400,
                1184,
                2609,
                3202,
                4583,
                6676,
                35526,
                2480,
                1296,
                22667,
                13,
                51372
            ],
            "temperature": 0.0,
            "avg_logprob": -0.333610381217713,
            "compression_ratio": 1.6482758620689655,
            "no_speech_prob": 0.00409698486328125,
            "words": [
                {
                    "start": 2170.45,
                    "end": 2170.89,
                    "word": " attention",
                    "probability": 0.61474609375
                },
                {
                    "start": 2170.89,
                    "end": 2171.41,
                    "word": " layers",
                    "probability": 0.8681640625
                },
                {
                    "start": 2171.41,
                    "end": 2172.05,
                    "word": ".",
                    "probability": 0.4794921875
                },
                {
                    "start": 2172.05,
                    "end": 2172.19,
                    "word": " And",
                    "probability": 0.86181640625
                },
                {
                    "start": 2172.19,
                    "end": 2172.53,
                    "word": " each",
                    "probability": 0.8603515625
                },
                {
                    "start": 2172.53,
                    "end": 2173.31,
                    "word": " individual",
                    "probability": 0.7822265625
                },
                {
                    "start": 2173.31,
                    "end": 2174.29,
                    "word": " attention",
                    "probability": 0.8583984375
                },
                {
                    "start": 2174.29,
                    "end": 2174.83,
                    "word": " layer",
                    "probability": 0.7734375
                },
                {
                    "start": 2174.83,
                    "end": 2176.25,
                    "word": " runs",
                    "probability": 0.398193359375
                },
                {
                    "start": 2176.25,
                    "end": 2177.07,
                    "word": " contextual",
                    "probability": 0.83154296875
                },
                {
                    "start": 2177.07,
                    "end": 2177.75,
                    "word": " relationship",
                    "probability": 0.47705078125
                },
                {
                    "start": 2177.75,
                    "end": 2178.33,
                    "word": " between",
                    "probability": 0.7158203125
                },
                {
                    "start": 2178.33,
                    "end": 2179.23,
                    "word": " tokens",
                    "probability": 0.73095703125
                },
                {
                    "start": 2179.23,
                    "end": 2179.77,
                    "word": ".",
                    "probability": 0.861328125
                }
            ]
        },
        {
            "id": 319,
            "seek": 213263,
            "start": 2180.89,
            "end": 2184.27,
            "text": " And the lighter diagram shows attention rate between",
            "tokens": [
                51448,
                400,
                264,
                11546,
                10686,
                3110,
                3202,
                3314,
                1296,
                51616
            ],
            "temperature": 0.0,
            "avg_logprob": -0.333610381217713,
            "compression_ratio": 1.6482758620689655,
            "no_speech_prob": 0.00409698486328125,
            "words": [
                {
                    "start": 2180.89,
                    "end": 2181.19,
                    "word": " And",
                    "probability": 0.86865234375
                },
                {
                    "start": 2181.19,
                    "end": 2181.73,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 2181.73,
                    "end": 2182.01,
                    "word": " lighter",
                    "probability": 0.322998046875
                },
                {
                    "start": 2182.01,
                    "end": 2182.51,
                    "word": " diagram",
                    "probability": 0.77978515625
                },
                {
                    "start": 2182.51,
                    "end": 2182.95,
                    "word": " shows",
                    "probability": 0.806640625
                },
                {
                    "start": 2182.95,
                    "end": 2183.51,
                    "word": " attention",
                    "probability": 0.8447265625
                },
                {
                    "start": 2183.51,
                    "end": 2183.77,
                    "word": " rate",
                    "probability": 0.59033203125
                },
                {
                    "start": 2183.77,
                    "end": 2184.27,
                    "word": " between",
                    "probability": 0.7587890625
                }
            ]
        },
        {
            "id": 320,
            "seek": 215323,
            "start": 2184.83,
            "end": 2194.23,
            "text": " tokens. The transfer token has multiple attention rate for other tokens.",
            "tokens": [
                50392,
                22667,
                13,
                440,
                5003,
                14862,
                575,
                3866,
                3202,
                3314,
                337,
                661,
                22667,
                13,
                50804
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24237350622812906,
            "compression_ratio": 1.4137931034482758,
            "no_speech_prob": 0.01763916015625,
            "words": [
                {
                    "start": 2184.83,
                    "end": 2185.37,
                    "word": " tokens",
                    "probability": 0.0264434814453125
                },
                {
                    "start": 2185.37,
                    "end": 2185.85,
                    "word": ".",
                    "probability": 0.6943359375
                },
                {
                    "start": 2185.85,
                    "end": 2186.27,
                    "word": " The",
                    "probability": 0.767578125
                },
                {
                    "start": 2186.27,
                    "end": 2186.73,
                    "word": " transfer",
                    "probability": 0.2322998046875
                },
                {
                    "start": 2186.73,
                    "end": 2187.43,
                    "word": " token",
                    "probability": 0.72119140625
                },
                {
                    "start": 2187.43,
                    "end": 2188.27,
                    "word": " has",
                    "probability": 0.888671875
                },
                {
                    "start": 2188.27,
                    "end": 2188.95,
                    "word": " multiple",
                    "probability": 0.76953125
                },
                {
                    "start": 2188.95,
                    "end": 2189.89,
                    "word": " attention",
                    "probability": 0.419189453125
                },
                {
                    "start": 2189.89,
                    "end": 2190.23,
                    "word": " rate",
                    "probability": 0.056732177734375
                },
                {
                    "start": 2190.23,
                    "end": 2191.65,
                    "word": " for",
                    "probability": 0.81787109375
                },
                {
                    "start": 2191.65,
                    "end": 2192.63,
                    "word": " other",
                    "probability": 0.77734375
                },
                {
                    "start": 2192.63,
                    "end": 2193.19,
                    "word": " tokens",
                    "probability": 0.755859375
                },
                {
                    "start": 2193.19,
                    "end": 2194.23,
                    "word": ".",
                    "probability": 0.94384765625
                }
            ]
        },
        {
            "id": 321,
            "seek": 215323,
            "start": 2197.24,
            "end": 2207.48,
            "text": " In our email data, we have many non-English emails, and also non-English emails can include",
            "tokens": [
                50900,
                682,
                527,
                3796,
                1412,
                11,
                321,
                362,
                867,
                2107,
                12,
                31254,
                1933,
                12524,
                11,
                293,
                611,
                2107,
                12,
                31254,
                1933,
                12524,
                393,
                4090,
                51400
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24237350622812906,
            "compression_ratio": 1.4137931034482758,
            "no_speech_prob": 0.01763916015625,
            "words": [
                {
                    "start": 2197.24,
                    "end": 2197.9,
                    "word": " In",
                    "probability": 0.89990234375
                },
                {
                    "start": 2197.9,
                    "end": 2198.2,
                    "word": " our",
                    "probability": 0.81201171875
                },
                {
                    "start": 2198.2,
                    "end": 2198.5,
                    "word": " email",
                    "probability": 0.89013671875
                },
                {
                    "start": 2198.5,
                    "end": 2198.92,
                    "word": " data",
                    "probability": 0.80615234375
                },
                {
                    "start": 2198.92,
                    "end": 2199.16,
                    "word": ",",
                    "probability": 0.52783203125
                },
                {
                    "start": 2199.16,
                    "end": 2199.4,
                    "word": " we",
                    "probability": 0.8984375
                },
                {
                    "start": 2199.4,
                    "end": 2201.86,
                    "word": " have",
                    "probability": 0.88232421875
                },
                {
                    "start": 2201.86,
                    "end": 2202.64,
                    "word": " many",
                    "probability": 0.82861328125
                },
                {
                    "start": 2202.64,
                    "end": 2202.98,
                    "word": " non",
                    "probability": 0.89453125
                },
                {
                    "start": 2202.98,
                    "end": 2203.58,
                    "word": "-English",
                    "probability": 0.916015625
                },
                {
                    "start": 2203.58,
                    "end": 2204.18,
                    "word": " emails",
                    "probability": 0.8076171875
                },
                {
                    "start": 2204.18,
                    "end": 2204.92,
                    "word": ",",
                    "probability": 0.1937255859375
                },
                {
                    "start": 2204.92,
                    "end": 2205.08,
                    "word": " and",
                    "probability": 0.86474609375
                },
                {
                    "start": 2205.08,
                    "end": 2205.46,
                    "word": " also",
                    "probability": 0.77099609375
                },
                {
                    "start": 2205.46,
                    "end": 2205.76,
                    "word": " non",
                    "probability": 0.859375
                },
                {
                    "start": 2205.76,
                    "end": 2206.16,
                    "word": "-English",
                    "probability": 0.9488932291666666
                },
                {
                    "start": 2206.16,
                    "end": 2206.6,
                    "word": " emails",
                    "probability": 0.8330078125
                },
                {
                    "start": 2206.6,
                    "end": 2206.88,
                    "word": " can",
                    "probability": 0.8916015625
                },
                {
                    "start": 2206.88,
                    "end": 2207.48,
                    "word": " include",
                    "probability": 0.7939453125
                }
            ]
        },
        {
            "id": 322,
            "seek": 217179,
            "start": 2207.48,
            "end": 2216.34,
            "text": " English words. Also, English emails can include non-English words.",
            "tokens": [
                50372,
                3669,
                2283,
                13,
                2743,
                11,
                3669,
                12524,
                393,
                4090,
                2107,
                12,
                31254,
                1933,
                2283,
                13,
                50792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2712890513241291,
            "compression_ratio": 1.375,
            "no_speech_prob": 0.04473876953125,
            "words": [
                {
                    "start": 2207.48,
                    "end": 2208.04,
                    "word": " English",
                    "probability": 0.2496337890625
                },
                {
                    "start": 2208.04,
                    "end": 2211.64,
                    "word": " words",
                    "probability": 0.0914306640625
                },
                {
                    "start": 2211.64,
                    "end": 2212.06,
                    "word": ".",
                    "probability": 0.2005615234375
                },
                {
                    "start": 2212.06,
                    "end": 2212.5,
                    "word": " Also",
                    "probability": 0.77978515625
                },
                {
                    "start": 2212.5,
                    "end": 2213.38,
                    "word": ",",
                    "probability": 0.50830078125
                },
                {
                    "start": 2213.38,
                    "end": 2213.74,
                    "word": " English",
                    "probability": 0.91015625
                },
                {
                    "start": 2213.74,
                    "end": 2214.1,
                    "word": " emails",
                    "probability": 0.58056640625
                },
                {
                    "start": 2214.1,
                    "end": 2214.48,
                    "word": " can",
                    "probability": 0.880859375
                },
                {
                    "start": 2214.48,
                    "end": 2214.86,
                    "word": " include",
                    "probability": 0.81640625
                },
                {
                    "start": 2214.86,
                    "end": 2215.18,
                    "word": " non",
                    "probability": 0.86181640625
                },
                {
                    "start": 2215.18,
                    "end": 2215.48,
                    "word": "-English",
                    "probability": 0.9423828125
                },
                {
                    "start": 2215.48,
                    "end": 2215.9,
                    "word": " words",
                    "probability": 0.669921875
                },
                {
                    "start": 2215.9,
                    "end": 2216.34,
                    "word": ".",
                    "probability": 0.94677734375
                }
            ]
        },
        {
            "id": 323,
            "seek": 217179,
            "start": 2217.36,
            "end": 2226.04,
            "text": " The non-English emails account for 25% of our total P9 and malicious emails.",
            "tokens": [
                50852,
                440,
                2107,
                12,
                31254,
                1933,
                12524,
                2696,
                337,
                3552,
                4,
                295,
                527,
                3217,
                430,
                24,
                293,
                33496,
                12524,
                13,
                51272
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2712890513241291,
            "compression_ratio": 1.375,
            "no_speech_prob": 0.04473876953125,
            "words": [
                {
                    "start": 2217.36,
                    "end": 2217.7,
                    "word": " The",
                    "probability": 0.479248046875
                },
                {
                    "start": 2217.7,
                    "end": 2218.66,
                    "word": " non",
                    "probability": 0.91455078125
                },
                {
                    "start": 2218.66,
                    "end": 2218.94,
                    "word": "-English",
                    "probability": 0.9501953125
                },
                {
                    "start": 2218.94,
                    "end": 2219.28,
                    "word": " emails",
                    "probability": 0.720703125
                },
                {
                    "start": 2219.28,
                    "end": 2219.84,
                    "word": " account",
                    "probability": 0.7119140625
                },
                {
                    "start": 2219.84,
                    "end": 2220.96,
                    "word": " for",
                    "probability": 0.884765625
                },
                {
                    "start": 2220.96,
                    "end": 2221.78,
                    "word": " 25",
                    "probability": 0.865234375
                },
                {
                    "start": 2221.78,
                    "end": 2222.3,
                    "word": "%",
                    "probability": 0.7705078125
                },
                {
                    "start": 2222.3,
                    "end": 2222.64,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 2222.64,
                    "end": 2223.02,
                    "word": " our",
                    "probability": 0.7451171875
                },
                {
                    "start": 2223.02,
                    "end": 2223.56,
                    "word": " total",
                    "probability": 0.701171875
                },
                {
                    "start": 2223.56,
                    "end": 2224.28,
                    "word": " P9",
                    "probability": 0.854736328125
                },
                {
                    "start": 2224.28,
                    "end": 2224.58,
                    "word": " and",
                    "probability": 0.5458984375
                },
                {
                    "start": 2224.58,
                    "end": 2225.06,
                    "word": " malicious",
                    "probability": 0.06982421875
                },
                {
                    "start": 2225.06,
                    "end": 2225.6,
                    "word": " emails",
                    "probability": 0.7529296875
                },
                {
                    "start": 2225.6,
                    "end": 2226.04,
                    "word": ".",
                    "probability": 0.73779296875
                }
            ]
        },
        {
            "id": 324,
            "seek": 220049,
            "start": 2226.04,
            "end": 2238.24,
            "text": " So, we needed to support a multilingual model which can recognize different languages.",
            "tokens": [
                50388,
                407,
                11,
                321,
                2978,
                281,
                1406,
                257,
                2120,
                38219,
                2316,
                597,
                393,
                5521,
                819,
                8650,
                13,
                50952
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25486681304994174,
            "compression_ratio": 1.5696969696969696,
            "no_speech_prob": 0.01158905029296875,
            "words": [
                {
                    "start": 2226.04,
                    "end": 2226.86,
                    "word": " So",
                    "probability": 0.389892578125
                },
                {
                    "start": 2226.86,
                    "end": 2227.46,
                    "word": ",",
                    "probability": 0.334228515625
                },
                {
                    "start": 2227.46,
                    "end": 2227.6,
                    "word": " we",
                    "probability": 0.87158203125
                },
                {
                    "start": 2227.6,
                    "end": 2227.9,
                    "word": " needed",
                    "probability": 0.487060546875
                },
                {
                    "start": 2227.9,
                    "end": 2228.16,
                    "word": " to",
                    "probability": 0.89599609375
                },
                {
                    "start": 2228.16,
                    "end": 2228.66,
                    "word": " support",
                    "probability": 0.9130859375
                },
                {
                    "start": 2230.02,
                    "end": 2230.58,
                    "word": " a multilingual",
                    "probability": 0.866455078125
                },
                {
                    "start": 2230.58,
                    "end": 2233.1,
                    "word": " model",
                    "probability": 0.75390625
                },
                {
                    "start": 2233.1,
                    "end": 2233.48,
                    "word": " which",
                    "probability": 0.69775390625
                },
                {
                    "start": 2233.48,
                    "end": 2233.92,
                    "word": " can",
                    "probability": 0.8935546875
                },
                {
                    "start": 2233.92,
                    "end": 2235.52,
                    "word": " recognize",
                    "probability": 0.6943359375
                },
                {
                    "start": 2235.52,
                    "end": 2237.18,
                    "word": " different",
                    "probability": 0.68505859375
                },
                {
                    "start": 2237.18,
                    "end": 2237.7,
                    "word": " languages",
                    "probability": 0.8603515625
                },
                {
                    "start": 2237.7,
                    "end": 2238.24,
                    "word": ".",
                    "probability": 0.90576171875
                }
            ]
        },
        {
            "id": 325,
            "seek": 220049,
            "start": 2241.17,
            "end": 2249.03,
            "text": " How we can support multilingual emails? The solution is BOT comes with two versions,",
            "tokens": [
                51048,
                1012,
                321,
                393,
                1406,
                2120,
                38219,
                12524,
                30,
                440,
                3827,
                307,
                363,
                5068,
                1487,
                365,
                732,
                9606,
                11,
                51384
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25486681304994174,
            "compression_ratio": 1.5696969696969696,
            "no_speech_prob": 0.01158905029296875,
            "words": [
                {
                    "start": 2241.17,
                    "end": 2241.75,
                    "word": " How",
                    "probability": 0.8349609375
                },
                {
                    "start": 2241.75,
                    "end": 2241.95,
                    "word": " we",
                    "probability": 0.5205078125
                },
                {
                    "start": 2241.95,
                    "end": 2242.25,
                    "word": " can",
                    "probability": 0.88427734375
                },
                {
                    "start": 2242.25,
                    "end": 2242.83,
                    "word": " support",
                    "probability": 0.93603515625
                },
                {
                    "start": 2242.83,
                    "end": 2243.49,
                    "word": " multilingual",
                    "probability": 0.8974609375
                },
                {
                    "start": 2243.49,
                    "end": 2244.09,
                    "word": " emails",
                    "probability": 0.492919921875
                },
                {
                    "start": 2244.09,
                    "end": 2244.83,
                    "word": "?",
                    "probability": 0.9453125
                },
                {
                    "start": 2244.83,
                    "end": 2245.09,
                    "word": " The",
                    "probability": 0.78125
                },
                {
                    "start": 2245.09,
                    "end": 2245.69,
                    "word": " solution",
                    "probability": 0.83447265625
                },
                {
                    "start": 2245.69,
                    "end": 2246.15,
                    "word": " is",
                    "probability": 0.89013671875
                },
                {
                    "start": 2246.15,
                    "end": 2246.71,
                    "word": " BOT",
                    "probability": 0.54632568359375
                },
                {
                    "start": 2246.71,
                    "end": 2247.07,
                    "word": " comes",
                    "probability": 0.8388671875
                },
                {
                    "start": 2247.07,
                    "end": 2247.39,
                    "word": " with",
                    "probability": 0.81591796875
                },
                {
                    "start": 2247.39,
                    "end": 2247.67,
                    "word": " two",
                    "probability": 0.75830078125
                },
                {
                    "start": 2247.67,
                    "end": 2248.29,
                    "word": " versions",
                    "probability": 0.78076171875
                },
                {
                    "start": 2248.29,
                    "end": 2249.03,
                    "word": ",",
                    "probability": 0.493408203125
                }
            ]
        },
        {
            "id": 326,
            "seek": 220049,
            "start": 2249.03,
            "end": 2256.49,
            "text": " English and multilingual BOT. The English version was pre-trained with the English text",
            "tokens": [
                51424,
                3669,
                293,
                2120,
                38219,
                363,
                5068,
                13,
                440,
                3669,
                3037,
                390,
                659,
                12,
                17227,
                2001,
                365,
                264,
                3669,
                2487,
                51816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25486681304994174,
            "compression_ratio": 1.5696969696969696,
            "no_speech_prob": 0.01158905029296875,
            "words": [
                {
                    "start": 2249.03,
                    "end": 2249.41,
                    "word": " English",
                    "probability": 0.89111328125
                },
                {
                    "start": 2249.41,
                    "end": 2249.93,
                    "word": " and",
                    "probability": 0.88427734375
                },
                {
                    "start": 2249.93,
                    "end": 2250.61,
                    "word": " multilingual",
                    "probability": 0.822998046875
                },
                {
                    "start": 2250.61,
                    "end": 2251.21,
                    "word": " BOT",
                    "probability": 0.927978515625
                },
                {
                    "start": 2251.21,
                    "end": 2251.91,
                    "word": ".",
                    "probability": 0.951171875
                },
                {
                    "start": 2251.91,
                    "end": 2252.47,
                    "word": " The",
                    "probability": 0.77294921875
                },
                {
                    "start": 2252.47,
                    "end": 2252.85,
                    "word": " English",
                    "probability": 0.916015625
                },
                {
                    "start": 2252.85,
                    "end": 2253.39,
                    "word": " version",
                    "probability": 0.6953125
                },
                {
                    "start": 2253.39,
                    "end": 2253.89,
                    "word": " was",
                    "probability": 0.88232421875
                },
                {
                    "start": 2253.89,
                    "end": 2254.23,
                    "word": " pre",
                    "probability": 0.814453125
                },
                {
                    "start": 2254.23,
                    "end": 2254.51,
                    "word": "-trained",
                    "probability": 0.9117838541666666
                },
                {
                    "start": 2254.51,
                    "end": 2254.87,
                    "word": " with",
                    "probability": 0.8095703125
                },
                {
                    "start": 2254.87,
                    "end": 2255.41,
                    "word": " the",
                    "probability": 0.42333984375
                },
                {
                    "start": 2255.41,
                    "end": 2255.97,
                    "word": " English",
                    "probability": 0.7978515625
                },
                {
                    "start": 2255.97,
                    "end": 2256.49,
                    "word": " text",
                    "probability": 0.669921875
                }
            ]
        },
        {
            "id": 327,
            "seek": 223049,
            "start": 2256.49,
            "end": 2264.35,
            "text": " datasets, including Wikipedia, and has 30,000 English tokens.",
            "tokens": [
                50380,
                42856,
                11,
                3009,
                28999,
                11,
                293,
                575,
                2217,
                11,
                1360,
                3669,
                22667,
                13,
                50736
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24866072109767368,
            "compression_ratio": 1.4723618090452262,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2256.49,
                    "end": 2257.59,
                    "word": " datasets",
                    "probability": 0.350341796875
                },
                {
                    "start": 2257.59,
                    "end": 2257.71,
                    "word": ",",
                    "probability": 0.2232666015625
                },
                {
                    "start": 2257.71,
                    "end": 2258.21,
                    "word": " including",
                    "probability": 0.79052734375
                },
                {
                    "start": 2258.21,
                    "end": 2258.95,
                    "word": " Wikipedia",
                    "probability": 0.77099609375
                },
                {
                    "start": 2258.95,
                    "end": 2260.61,
                    "word": ",",
                    "probability": 0.505859375
                },
                {
                    "start": 2260.61,
                    "end": 2260.87,
                    "word": " and",
                    "probability": 0.8505859375
                },
                {
                    "start": 2260.87,
                    "end": 2261.87,
                    "word": " has",
                    "probability": 0.8173828125
                },
                {
                    "start": 2261.87,
                    "end": 2262.45,
                    "word": " 30",
                    "probability": 0.822265625
                },
                {
                    "start": 2262.45,
                    "end": 2263.05,
                    "word": ",000",
                    "probability": 0.951416015625
                },
                {
                    "start": 2263.05,
                    "end": 2263.47,
                    "word": " English",
                    "probability": 0.896484375
                },
                {
                    "start": 2263.47,
                    "end": 2263.89,
                    "word": " tokens",
                    "probability": 0.6689453125
                },
                {
                    "start": 2263.89,
                    "end": 2264.35,
                    "word": ".",
                    "probability": 0.95849609375
                }
            ]
        },
        {
            "id": 328,
            "seek": 223049,
            "start": 2266.21,
            "end": 2274.37,
            "text": " The multilingual version was pre-trained with large text datasets from more than 100 languages.",
            "tokens": [
                50852,
                440,
                2120,
                38219,
                3037,
                390,
                659,
                12,
                17227,
                2001,
                365,
                2416,
                2487,
                42856,
                490,
                544,
                813,
                2319,
                8650,
                13,
                51240
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24866072109767368,
            "compression_ratio": 1.4723618090452262,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2266.21,
                    "end": 2266.67,
                    "word": " The",
                    "probability": 0.79150390625
                },
                {
                    "start": 2266.67,
                    "end": 2267.31,
                    "word": " multilingual",
                    "probability": 0.8408203125
                },
                {
                    "start": 2267.31,
                    "end": 2267.79,
                    "word": " version",
                    "probability": 0.71923828125
                },
                {
                    "start": 2267.79,
                    "end": 2268.45,
                    "word": " was",
                    "probability": 0.88818359375
                },
                {
                    "start": 2268.45,
                    "end": 2268.81,
                    "word": " pre",
                    "probability": 0.80712890625
                },
                {
                    "start": 2268.81,
                    "end": 2269.17,
                    "word": "-trained",
                    "probability": 0.9085286458333334
                },
                {
                    "start": 2269.17,
                    "end": 2269.49,
                    "word": " with",
                    "probability": 0.8134765625
                },
                {
                    "start": 2269.49,
                    "end": 2270.01,
                    "word": " large",
                    "probability": 0.89794921875
                },
                {
                    "start": 2270.01,
                    "end": 2270.39,
                    "word": " text",
                    "probability": 0.8193359375
                },
                {
                    "start": 2270.39,
                    "end": 2270.91,
                    "word": " datasets",
                    "probability": 0.82421875
                },
                {
                    "start": 2270.91,
                    "end": 2272.11,
                    "word": " from",
                    "probability": 0.7568359375
                },
                {
                    "start": 2272.11,
                    "end": 2272.43,
                    "word": " more",
                    "probability": 0.8623046875
                },
                {
                    "start": 2272.43,
                    "end": 2272.67,
                    "word": " than",
                    "probability": 0.888671875
                },
                {
                    "start": 2272.67,
                    "end": 2273.27,
                    "word": " 100",
                    "probability": 0.857421875
                },
                {
                    "start": 2273.27,
                    "end": 2273.91,
                    "word": " languages",
                    "probability": 0.888671875
                },
                {
                    "start": 2273.91,
                    "end": 2274.37,
                    "word": ".",
                    "probability": 0.455322265625
                }
            ]
        },
        {
            "id": 329,
            "seek": 223049,
            "start": 2274.97,
            "end": 2280.67,
            "text": " And this version has four times large vocabulary size, which is 120 tokens,",
            "tokens": [
                51280,
                400,
                341,
                3037,
                575,
                1451,
                1413,
                2416,
                19864,
                2744,
                11,
                597,
                307,
                10411,
                22667,
                11,
                51548
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24866072109767368,
            "compression_ratio": 1.4723618090452262,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2274.97,
                    "end": 2275.27,
                    "word": " And",
                    "probability": 0.493896484375
                },
                {
                    "start": 2275.27,
                    "end": 2275.71,
                    "word": " this",
                    "probability": 0.84423828125
                },
                {
                    "start": 2275.71,
                    "end": 2276.13,
                    "word": " version",
                    "probability": 0.72998046875
                },
                {
                    "start": 2276.13,
                    "end": 2276.55,
                    "word": " has",
                    "probability": 0.888671875
                },
                {
                    "start": 2276.55,
                    "end": 2276.81,
                    "word": " four",
                    "probability": 0.34765625
                },
                {
                    "start": 2276.81,
                    "end": 2277.11,
                    "word": " times",
                    "probability": 0.802734375
                },
                {
                    "start": 2277.11,
                    "end": 2277.49,
                    "word": " large",
                    "probability": 0.414794921875
                },
                {
                    "start": 2277.49,
                    "end": 2277.93,
                    "word": " vocabulary",
                    "probability": 0.58349609375
                },
                {
                    "start": 2277.93,
                    "end": 2278.35,
                    "word": " size",
                    "probability": 0.7099609375
                },
                {
                    "start": 2278.35,
                    "end": 2278.41,
                    "word": ",",
                    "probability": 0.81201171875
                },
                {
                    "start": 2278.41,
                    "end": 2278.63,
                    "word": " which",
                    "probability": 0.90625
                },
                {
                    "start": 2278.63,
                    "end": 2278.79,
                    "word": " is",
                    "probability": 0.890625
                },
                {
                    "start": 2278.79,
                    "end": 2279.53,
                    "word": " 120",
                    "probability": 0.845703125
                },
                {
                    "start": 2279.53,
                    "end": 2280.21,
                    "word": " tokens",
                    "probability": 0.7431640625
                },
                {
                    "start": 2280.21,
                    "end": 2280.67,
                    "word": ",",
                    "probability": 0.869140625
                }
            ]
        },
        {
            "id": 330,
            "seek": 223049,
            "start": 2280.67,
            "end": 2285.79,
            "text": " which will cover many Unicode characters and Unicode words.",
            "tokens": [
                51592,
                597,
                486,
                2060,
                867,
                1156,
                299,
                1429,
                4342,
                293,
                1156,
                299,
                1429,
                2283,
                13,
                51812
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24866072109767368,
            "compression_ratio": 1.4723618090452262,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2280.67,
                    "end": 2281.47,
                    "word": " which",
                    "probability": 0.91650390625
                },
                {
                    "start": 2281.47,
                    "end": 2281.69,
                    "word": " will",
                    "probability": 0.7587890625
                },
                {
                    "start": 2281.69,
                    "end": 2281.97,
                    "word": " cover",
                    "probability": 0.85595703125
                },
                {
                    "start": 2281.97,
                    "end": 2282.37,
                    "word": " many",
                    "probability": 0.826171875
                },
                {
                    "start": 2282.37,
                    "end": 2282.95,
                    "word": " Unicode",
                    "probability": 0.9303385416666666
                },
                {
                    "start": 2282.95,
                    "end": 2283.69,
                    "word": " characters",
                    "probability": 0.72412109375
                },
                {
                    "start": 2283.69,
                    "end": 2284.15,
                    "word": " and",
                    "probability": 0.89013671875
                },
                {
                    "start": 2284.15,
                    "end": 2285.07,
                    "word": " Unicode",
                    "probability": 0.96142578125
                },
                {
                    "start": 2285.07,
                    "end": 2285.53,
                    "word": " words",
                    "probability": 0.77099609375
                },
                {
                    "start": 2285.53,
                    "end": 2285.79,
                    "word": ".",
                    "probability": 0.99658203125
                }
            ]
        },
        {
            "id": 331,
            "seek": 225693,
            "start": 2286.49,
            "end": 2294.89,
            "text": " So, we fine-tuned a multilingual BOT for our multilingual emails.",
            "tokens": [
                50380,
                407,
                11,
                321,
                2489,
                12,
                83,
                43703,
                257,
                2120,
                38219,
                363,
                5068,
                337,
                527,
                2120,
                38219,
                12524,
                13,
                50764
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22788664547063536,
            "compression_ratio": 1.551948051948052,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 2286.49,
                    "end": 2287.23,
                    "word": " So",
                    "probability": 0.3662109375
                },
                {
                    "start": 2287.23,
                    "end": 2287.69,
                    "word": ",",
                    "probability": 0.396240234375
                },
                {
                    "start": 2287.69,
                    "end": 2288.23,
                    "word": " we",
                    "probability": 0.8701171875
                },
                {
                    "start": 2288.23,
                    "end": 2288.63,
                    "word": " fine",
                    "probability": 0.86083984375
                },
                {
                    "start": 2288.63,
                    "end": 2288.97,
                    "word": "-tuned",
                    "probability": 0.8782552083333334
                },
                {
                    "start": 2289.33,
                    "end": 2289.97,
                    "word": " a multilingual",
                    "probability": 0.864013671875
                },
                {
                    "start": 2289.97,
                    "end": 2291.13,
                    "word": " BOT",
                    "probability": 0.4469108581542969
                },
                {
                    "start": 2291.13,
                    "end": 2292.29,
                    "word": " for",
                    "probability": 0.84375
                },
                {
                    "start": 2292.29,
                    "end": 2293.15,
                    "word": " our",
                    "probability": 0.6689453125
                },
                {
                    "start": 2293.15,
                    "end": 2293.89,
                    "word": " multilingual",
                    "probability": 0.906005859375
                },
                {
                    "start": 2293.89,
                    "end": 2294.43,
                    "word": " emails",
                    "probability": 0.72314453125
                },
                {
                    "start": 2294.43,
                    "end": 2294.89,
                    "word": ".",
                    "probability": 0.90673828125
                }
            ]
        },
        {
            "id": 332,
            "seek": 225693,
            "start": 2298.27,
            "end": 2306.19,
            "text": " The second set of results from email headers. We can extract multiple indicators from",
            "tokens": [
                50856,
                440,
                1150,
                992,
                295,
                3542,
                490,
                3796,
                45101,
                13,
                492,
                393,
                8947,
                3866,
                22176,
                490,
                51240
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22788664547063536,
            "compression_ratio": 1.551948051948052,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 2298.27,
                    "end": 2298.73,
                    "word": " The",
                    "probability": 0.75341796875
                },
                {
                    "start": 2298.73,
                    "end": 2298.97,
                    "word": " second",
                    "probability": 0.77978515625
                },
                {
                    "start": 2298.97,
                    "end": 2299.23,
                    "word": " set",
                    "probability": 0.89306640625
                },
                {
                    "start": 2299.23,
                    "end": 2299.37,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 2299.37,
                    "end": 2299.61,
                    "word": " results",
                    "probability": 0.2021484375
                },
                {
                    "start": 2299.61,
                    "end": 2300.07,
                    "word": " from",
                    "probability": 0.58740234375
                },
                {
                    "start": 2300.07,
                    "end": 2300.51,
                    "word": " email",
                    "probability": 0.837890625
                },
                {
                    "start": 2300.51,
                    "end": 2300.93,
                    "word": " headers",
                    "probability": 0.9130859375
                },
                {
                    "start": 2300.93,
                    "end": 2301.85,
                    "word": ".",
                    "probability": 0.70458984375
                },
                {
                    "start": 2301.89,
                    "end": 2302.45,
                    "word": " We",
                    "probability": 0.8701171875
                },
                {
                    "start": 2302.45,
                    "end": 2302.89,
                    "word": " can",
                    "probability": 0.8876953125
                },
                {
                    "start": 2302.89,
                    "end": 2304.15,
                    "word": " extract",
                    "probability": 0.75244140625
                },
                {
                    "start": 2304.15,
                    "end": 2304.79,
                    "word": " multiple",
                    "probability": 0.810546875
                },
                {
                    "start": 2304.79,
                    "end": 2305.59,
                    "word": " indicators",
                    "probability": 0.9091796875
                },
                {
                    "start": 2305.59,
                    "end": 2306.19,
                    "word": " from",
                    "probability": 0.8134765625
                }
            ]
        },
        {
            "id": 333,
            "seek": 225693,
            "start": 2307.17,
            "end": 2314.97,
            "text": " email header builders. For example, the first one, we check whether the emails are from",
            "tokens": [
                51296,
                3796,
                23117,
                36281,
                13,
                1171,
                1365,
                11,
                264,
                700,
                472,
                11,
                321,
                1520,
                1968,
                264,
                12524,
                366,
                490,
                51676
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22788664547063536,
            "compression_ratio": 1.551948051948052,
            "no_speech_prob": 0.004608154296875,
            "words": [
                {
                    "start": 2307.17,
                    "end": 2307.73,
                    "word": " email",
                    "probability": 0.8828125
                },
                {
                    "start": 2307.73,
                    "end": 2308.49,
                    "word": " header",
                    "probability": 0.564453125
                },
                {
                    "start": 2308.49,
                    "end": 2308.95,
                    "word": " builders",
                    "probability": 0.50439453125
                },
                {
                    "start": 2308.95,
                    "end": 2309.41,
                    "word": ".",
                    "probability": 0.89697265625
                },
                {
                    "start": 2309.41,
                    "end": 2309.57,
                    "word": " For",
                    "probability": 0.89404296875
                },
                {
                    "start": 2309.57,
                    "end": 2309.99,
                    "word": " example",
                    "probability": 0.92919921875
                },
                {
                    "start": 2309.99,
                    "end": 2310.29,
                    "word": ",",
                    "probability": 0.955078125
                },
                {
                    "start": 2310.29,
                    "end": 2310.51,
                    "word": " the",
                    "probability": 0.69287109375
                },
                {
                    "start": 2310.51,
                    "end": 2310.81,
                    "word": " first",
                    "probability": 0.7919921875
                },
                {
                    "start": 2310.81,
                    "end": 2311.23,
                    "word": " one",
                    "probability": 0.8134765625
                },
                {
                    "start": 2311.23,
                    "end": 2312.05,
                    "word": ",",
                    "probability": 0.52978515625
                },
                {
                    "start": 2312.05,
                    "end": 2312.47,
                    "word": " we",
                    "probability": 0.8701171875
                },
                {
                    "start": 2312.47,
                    "end": 2313.15,
                    "word": " check",
                    "probability": 0.76123046875
                },
                {
                    "start": 2313.15,
                    "end": 2313.45,
                    "word": " whether",
                    "probability": 0.84375
                },
                {
                    "start": 2313.45,
                    "end": 2313.75,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 2313.75,
                    "end": 2314.21,
                    "word": " emails",
                    "probability": 0.75146484375
                },
                {
                    "start": 2314.21,
                    "end": 2314.57,
                    "word": " are",
                    "probability": 0.890625
                },
                {
                    "start": 2314.57,
                    "end": 2314.97,
                    "word": " from",
                    "probability": 0.80859375
                }
            ]
        },
        {
            "id": 334,
            "seek": 228429,
            "start": 2314.97,
            "end": 2325.09,
            "text": " internal or external. We can compare the domain of recipient and sender. For example, here,",
            "tokens": [
                50380,
                6920,
                420,
                8320,
                13,
                492,
                393,
                6794,
                264,
                9274,
                295,
                26216,
                293,
                2845,
                260,
                13,
                1171,
                1365,
                11,
                510,
                11,
                50812
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2832341354990762,
            "compression_ratio": 1.5272727272727273,
            "no_speech_prob": 0.00782012939453125,
            "words": [
                {
                    "start": 2314.97,
                    "end": 2315.81,
                    "word": " internal",
                    "probability": 0.043914794921875
                },
                {
                    "start": 2315.81,
                    "end": 2316.09,
                    "word": " or",
                    "probability": 0.71435546875
                },
                {
                    "start": 2316.09,
                    "end": 2316.57,
                    "word": " external",
                    "probability": 0.79443359375
                },
                {
                    "start": 2316.57,
                    "end": 2316.87,
                    "word": ".",
                    "probability": 0.112060546875
                },
                {
                    "start": 2316.87,
                    "end": 2317.47,
                    "word": " We",
                    "probability": 0.837890625
                },
                {
                    "start": 2317.47,
                    "end": 2317.67,
                    "word": " can",
                    "probability": 0.8974609375
                },
                {
                    "start": 2317.67,
                    "end": 2318.11,
                    "word": " compare",
                    "probability": 0.8388671875
                },
                {
                    "start": 2318.11,
                    "end": 2318.37,
                    "word": " the",
                    "probability": 0.7958984375
                },
                {
                    "start": 2318.37,
                    "end": 2318.85,
                    "word": " domain",
                    "probability": 0.9091796875
                },
                {
                    "start": 2318.85,
                    "end": 2319.77,
                    "word": " of",
                    "probability": 0.80810546875
                },
                {
                    "start": 2319.77,
                    "end": 2320.55,
                    "word": " recipient",
                    "probability": 0.67578125
                },
                {
                    "start": 2320.55,
                    "end": 2321.71,
                    "word": " and",
                    "probability": 0.88134765625
                },
                {
                    "start": 2321.71,
                    "end": 2322.99,
                    "word": " sender",
                    "probability": 0.784423828125
                },
                {
                    "start": 2322.99,
                    "end": 2323.07,
                    "word": ".",
                    "probability": 0.8095703125
                },
                {
                    "start": 2323.07,
                    "end": 2323.17,
                    "word": " For",
                    "probability": 0.86572265625
                },
                {
                    "start": 2323.17,
                    "end": 2323.53,
                    "word": " example",
                    "probability": 0.92724609375
                },
                {
                    "start": 2323.53,
                    "end": 2323.63,
                    "word": ",",
                    "probability": 0.5380859375
                },
                {
                    "start": 2323.63,
                    "end": 2323.99,
                    "word": " here",
                    "probability": 0.7607421875
                },
                {
                    "start": 2323.99,
                    "end": 2325.09,
                    "word": ",",
                    "probability": 0.265625
                }
            ]
        },
        {
            "id": 335,
            "seek": 228429,
            "start": 2325.58,
            "end": 2332.75,
            "text": " the domain name is a similar looking one, but actually there are extra ads.",
            "tokens": [
                50896,
                264,
                9274,
                1315,
                307,
                257,
                2531,
                1237,
                472,
                11,
                457,
                767,
                456,
                366,
                2857,
                10342,
                13,
                51244
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2832341354990762,
            "compression_ratio": 1.5272727272727273,
            "no_speech_prob": 0.00782012939453125,
            "words": [
                {
                    "start": 2325.58,
                    "end": 2326.01,
                    "word": " the",
                    "probability": 0.78564453125
                },
                {
                    "start": 2326.01,
                    "end": 2326.49,
                    "word": " domain",
                    "probability": 0.888671875
                },
                {
                    "start": 2326.49,
                    "end": 2326.79,
                    "word": " name",
                    "probability": 0.8095703125
                },
                {
                    "start": 2326.79,
                    "end": 2326.95,
                    "word": " is",
                    "probability": 0.884765625
                },
                {
                    "start": 2327.35,
                    "end": 2329.59,
                    "word": " a similar",
                    "probability": 0.54052734375
                },
                {
                    "start": 2329.59,
                    "end": 2330.07,
                    "word": " looking",
                    "probability": 0.6640625
                },
                {
                    "start": 2330.07,
                    "end": 2330.25,
                    "word": " one",
                    "probability": 0.64501953125
                },
                {
                    "start": 2330.25,
                    "end": 2330.33,
                    "word": ",",
                    "probability": 0.5986328125
                },
                {
                    "start": 2330.33,
                    "end": 2330.45,
                    "word": " but",
                    "probability": 0.88037109375
                },
                {
                    "start": 2330.45,
                    "end": 2330.77,
                    "word": " actually",
                    "probability": 0.724609375
                },
                {
                    "start": 2330.77,
                    "end": 2330.99,
                    "word": " there",
                    "probability": 0.68505859375
                },
                {
                    "start": 2330.99,
                    "end": 2331.09,
                    "word": " are",
                    "probability": 0.496826171875
                },
                {
                    "start": 2331.09,
                    "end": 2331.71,
                    "word": " extra",
                    "probability": 0.71826171875
                },
                {
                    "start": 2331.71,
                    "end": 2332.21,
                    "word": " ads",
                    "probability": 0.0948486328125
                },
                {
                    "start": 2332.21,
                    "end": 2332.75,
                    "word": ".",
                    "probability": 0.65576171875
                }
            ]
        },
        {
            "id": 336,
            "seek": 228429,
            "start": 2332.75,
            "end": 2342.33,
            "text": " So, this one we consider as external email. And then, we can also use external reply",
            "tokens": [
                51244,
                407,
                11,
                341,
                472,
                321,
                1949,
                382,
                8320,
                3796,
                13,
                400,
                550,
                11,
                321,
                393,
                611,
                764,
                8320,
                16972,
                51728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2832341354990762,
            "compression_ratio": 1.5272727272727273,
            "no_speech_prob": 0.00782012939453125,
            "words": [
                {
                    "start": 2332.75,
                    "end": 2332.89,
                    "word": " So",
                    "probability": 0.83740234375
                },
                {
                    "start": 2332.89,
                    "end": 2333.05,
                    "word": ",",
                    "probability": 0.123291015625
                },
                {
                    "start": 2333.05,
                    "end": 2333.23,
                    "word": " this",
                    "probability": 0.8486328125
                },
                {
                    "start": 2333.23,
                    "end": 2333.51,
                    "word": " one",
                    "probability": 0.81591796875
                },
                {
                    "start": 2333.51,
                    "end": 2333.73,
                    "word": " we",
                    "probability": 0.7578125
                },
                {
                    "start": 2333.73,
                    "end": 2334.17,
                    "word": " consider",
                    "probability": 0.7275390625
                },
                {
                    "start": 2334.17,
                    "end": 2334.65,
                    "word": " as",
                    "probability": 0.84619140625
                },
                {
                    "start": 2334.65,
                    "end": 2335.21,
                    "word": " external",
                    "probability": 0.71484375
                },
                {
                    "start": 2335.21,
                    "end": 2336.23,
                    "word": " email",
                    "probability": 0.92919921875
                },
                {
                    "start": 2336.23,
                    "end": 2337.09,
                    "word": ".",
                    "probability": 0.85791015625
                },
                {
                    "start": 2337.65,
                    "end": 2338.07,
                    "word": " And",
                    "probability": 0.798828125
                },
                {
                    "start": 2338.07,
                    "end": 2338.55,
                    "word": " then",
                    "probability": 0.6982421875
                },
                {
                    "start": 2338.55,
                    "end": 2338.71,
                    "word": ",",
                    "probability": 0.412109375
                },
                {
                    "start": 2338.71,
                    "end": 2339.21,
                    "word": " we",
                    "probability": 0.8837890625
                },
                {
                    "start": 2339.21,
                    "end": 2339.41,
                    "word": " can",
                    "probability": 0.8984375
                },
                {
                    "start": 2339.41,
                    "end": 2339.67,
                    "word": " also",
                    "probability": 0.82373046875
                },
                {
                    "start": 2339.67,
                    "end": 2340.63,
                    "word": " use",
                    "probability": 0.8212890625
                },
                {
                    "start": 2340.63,
                    "end": 2341.77,
                    "word": " external",
                    "probability": 0.646484375
                },
                {
                    "start": 2341.77,
                    "end": 2342.33,
                    "word": " reply",
                    "probability": 0.74267578125
                }
            ]
        },
        {
            "id": 337,
            "seek": 230967,
            "start": 2342.99,
            "end": 2352.23,
            "text": " by comparing the domain of from and reply to. Often, targeted phishing attackers use",
            "tokens": [
                50400,
                538,
                15763,
                264,
                9274,
                295,
                490,
                293,
                16972,
                281,
                13,
                20043,
                11,
                15045,
                903,
                3807,
                45129,
                764,
                50852
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30164196127552095,
            "compression_ratio": 1.587878787878788,
            "no_speech_prob": 0.01250457763671875,
            "words": [
                {
                    "start": 2342.99,
                    "end": 2343.37,
                    "word": " by",
                    "probability": 0.54541015625
                },
                {
                    "start": 2343.37,
                    "end": 2343.97,
                    "word": " comparing",
                    "probability": 0.76708984375
                },
                {
                    "start": 2343.97,
                    "end": 2344.33,
                    "word": " the",
                    "probability": 0.76806640625
                },
                {
                    "start": 2344.33,
                    "end": 2344.73,
                    "word": " domain",
                    "probability": 0.82958984375
                },
                {
                    "start": 2344.73,
                    "end": 2344.97,
                    "word": " of",
                    "probability": 0.83251953125
                },
                {
                    "start": 2344.97,
                    "end": 2345.45,
                    "word": " from",
                    "probability": 0.0229949951171875
                },
                {
                    "start": 2345.45,
                    "end": 2345.77,
                    "word": " and",
                    "probability": 0.86767578125
                },
                {
                    "start": 2345.77,
                    "end": 2346.15,
                    "word": " reply",
                    "probability": 0.771484375
                },
                {
                    "start": 2346.15,
                    "end": 2346.57,
                    "word": " to",
                    "probability": 0.45703125
                },
                {
                    "start": 2346.57,
                    "end": 2347.33,
                    "word": ".",
                    "probability": 0.91357421875
                },
                {
                    "start": 2347.57,
                    "end": 2347.93,
                    "word": " Often",
                    "probability": 0.74365234375
                },
                {
                    "start": 2347.93,
                    "end": 2349.01,
                    "word": ",",
                    "probability": 0.408935546875
                },
                {
                    "start": 2349.01,
                    "end": 2349.93,
                    "word": " targeted",
                    "probability": 0.7607421875
                },
                {
                    "start": 2349.93,
                    "end": 2350.33,
                    "word": " phishing",
                    "probability": 0.856689453125
                },
                {
                    "start": 2350.33,
                    "end": 2351.53,
                    "word": " attackers",
                    "probability": 0.75048828125
                },
                {
                    "start": 2351.53,
                    "end": 2352.23,
                    "word": " use",
                    "probability": 0.76171875
                }
            ]
        },
        {
            "id": 338,
            "seek": 230967,
            "start": 2353.05,
            "end": 2360.09,
            "text": " a different domain for reply to. And also, we collect the size of the recipients and the size",
            "tokens": [
                50884,
                257,
                819,
                9274,
                337,
                16972,
                281,
                13,
                400,
                611,
                11,
                321,
                2500,
                264,
                2744,
                295,
                264,
                32440,
                293,
                264,
                2744,
                51248
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30164196127552095,
            "compression_ratio": 1.587878787878788,
            "no_speech_prob": 0.01250457763671875,
            "words": [
                {
                    "start": 2353.05,
                    "end": 2353.79,
                    "word": " a different",
                    "probability": 0.673828125
                },
                {
                    "start": 2353.79,
                    "end": 2354.37,
                    "word": " domain",
                    "probability": 0.884765625
                },
                {
                    "start": 2354.37,
                    "end": 2354.63,
                    "word": " for",
                    "probability": 0.85791015625
                },
                {
                    "start": 2354.63,
                    "end": 2354.97,
                    "word": " reply",
                    "probability": 0.76220703125
                },
                {
                    "start": 2354.97,
                    "end": 2355.33,
                    "word": " to",
                    "probability": 0.87744140625
                },
                {
                    "start": 2355.33,
                    "end": 2356.09,
                    "word": ".",
                    "probability": 0.748046875
                },
                {
                    "start": 2356.11,
                    "end": 2356.27,
                    "word": " And",
                    "probability": 0.70068359375
                },
                {
                    "start": 2356.27,
                    "end": 2356.55,
                    "word": " also",
                    "probability": 0.79248046875
                },
                {
                    "start": 2356.55,
                    "end": 2356.67,
                    "word": ",",
                    "probability": 0.4638671875
                },
                {
                    "start": 2356.67,
                    "end": 2356.79,
                    "word": " we",
                    "probability": 0.89892578125
                },
                {
                    "start": 2356.79,
                    "end": 2357.85,
                    "word": " collect",
                    "probability": 0.75244140625
                },
                {
                    "start": 2357.85,
                    "end": 2358.09,
                    "word": " the",
                    "probability": 0.392333984375
                },
                {
                    "start": 2358.09,
                    "end": 2358.35,
                    "word": " size",
                    "probability": 0.80810546875
                },
                {
                    "start": 2358.35,
                    "end": 2358.57,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 2358.57,
                    "end": 2358.65,
                    "word": " the",
                    "probability": 0.1708984375
                },
                {
                    "start": 2358.65,
                    "end": 2359.15,
                    "word": " recipients",
                    "probability": 0.55517578125
                },
                {
                    "start": 2359.15,
                    "end": 2359.65,
                    "word": " and",
                    "probability": 0.80859375
                },
                {
                    "start": 2359.65,
                    "end": 2359.81,
                    "word": " the",
                    "probability": 0.76123046875
                },
                {
                    "start": 2359.81,
                    "end": 2360.09,
                    "word": " size",
                    "probability": 0.77783203125
                }
            ]
        },
        {
            "id": 339,
            "seek": 230967,
            "start": 2360.09,
            "end": 2367.71,
            "text": " of the carbon copy recipients as additional indicators. It is obvious that targeted",
            "tokens": [
                51248,
                295,
                264,
                5954,
                5055,
                32440,
                382,
                4497,
                22176,
                13,
                467,
                307,
                6322,
                300,
                15045,
                51636
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30164196127552095,
            "compression_ratio": 1.587878787878788,
            "no_speech_prob": 0.01250457763671875,
            "words": [
                {
                    "start": 2360.09,
                    "end": 2360.27,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 2360.27,
                    "end": 2360.55,
                    "word": " the",
                    "probability": 0.62158203125
                },
                {
                    "start": 2360.55,
                    "end": 2360.87,
                    "word": " carbon",
                    "probability": 0.779296875
                },
                {
                    "start": 2360.87,
                    "end": 2361.25,
                    "word": " copy",
                    "probability": 0.66357421875
                },
                {
                    "start": 2361.25,
                    "end": 2361.99,
                    "word": " recipients",
                    "probability": 0.64697265625
                },
                {
                    "start": 2361.99,
                    "end": 2362.33,
                    "word": " as",
                    "probability": 0.5771484375
                },
                {
                    "start": 2362.33,
                    "end": 2362.97,
                    "word": " additional",
                    "probability": 0.84619140625
                },
                {
                    "start": 2362.97,
                    "end": 2363.89,
                    "word": " indicators",
                    "probability": 0.88232421875
                },
                {
                    "start": 2363.89,
                    "end": 2364.65,
                    "word": ".",
                    "probability": 0.9853515625
                },
                {
                    "start": 2365.11,
                    "end": 2365.41,
                    "word": " It",
                    "probability": 0.89892578125
                },
                {
                    "start": 2365.41,
                    "end": 2365.57,
                    "word": " is",
                    "probability": 0.8994140625
                },
                {
                    "start": 2365.57,
                    "end": 2365.95,
                    "word": " obvious",
                    "probability": 0.8955078125
                },
                {
                    "start": 2365.95,
                    "end": 2366.39,
                    "word": " that",
                    "probability": 0.87451171875
                },
                {
                    "start": 2366.39,
                    "end": 2367.71,
                    "word": " targeted",
                    "probability": 0.77783203125
                }
            ]
        },
        {
            "id": 340,
            "seek": 233687,
            "start": 2367.71,
            "end": 2375.21,
            "text": " phishing attacks will have only single recipient. However, many user",
            "tokens": [
                50368,
                903,
                3807,
                8122,
                486,
                362,
                787,
                2167,
                26216,
                13,
                2908,
                11,
                867,
                4195,
                50724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2709134615384615,
            "compression_ratio": 1.7430555555555556,
            "no_speech_prob": 0.067138671875,
            "words": [
                {
                    "start": 2367.71,
                    "end": 2368.21,
                    "word": " phishing",
                    "probability": 0.4274749755859375
                },
                {
                    "start": 2368.21,
                    "end": 2368.57,
                    "word": " attacks",
                    "probability": 0.71435546875
                },
                {
                    "start": 2368.57,
                    "end": 2368.97,
                    "word": " will",
                    "probability": 0.697265625
                },
                {
                    "start": 2368.97,
                    "end": 2369.71,
                    "word": " have",
                    "probability": 0.84375
                },
                {
                    "start": 2369.71,
                    "end": 2370.07,
                    "word": " only",
                    "probability": 0.8125
                },
                {
                    "start": 2370.07,
                    "end": 2371.07,
                    "word": " single",
                    "probability": 0.72998046875
                },
                {
                    "start": 2371.07,
                    "end": 2371.99,
                    "word": " recipient",
                    "probability": 0.5166015625
                },
                {
                    "start": 2371.99,
                    "end": 2372.71,
                    "word": ".",
                    "probability": 0.63525390625
                },
                {
                    "start": 2372.71,
                    "end": 2373.05,
                    "word": " However",
                    "probability": 0.84033203125
                },
                {
                    "start": 2373.05,
                    "end": 2373.27,
                    "word": ",",
                    "probability": 0.72119140625
                },
                {
                    "start": 2373.27,
                    "end": 2373.49,
                    "word": " many",
                    "probability": 0.80810546875
                },
                {
                    "start": 2373.49,
                    "end": 2375.21,
                    "word": " user",
                    "probability": 0.7802734375
                }
            ]
        },
        {
            "id": 341,
            "seek": 233687,
            "start": 2377.11,
            "end": 2385.09,
            "text": " phishing emails will have multiple recipients or carbon copy recipients.",
            "tokens": [
                50852,
                903,
                3807,
                12524,
                486,
                362,
                3866,
                32440,
                420,
                5954,
                5055,
                32440,
                13,
                51188
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2709134615384615,
            "compression_ratio": 1.7430555555555556,
            "no_speech_prob": 0.067138671875,
            "words": [
                {
                    "start": 2377.11,
                    "end": 2377.91,
                    "word": " phishing",
                    "probability": 0.67333984375
                },
                {
                    "start": 2377.91,
                    "end": 2378.57,
                    "word": " emails",
                    "probability": 0.7109375
                },
                {
                    "start": 2378.57,
                    "end": 2379.37,
                    "word": " will",
                    "probability": 0.78955078125
                },
                {
                    "start": 2379.37,
                    "end": 2379.73,
                    "word": " have",
                    "probability": 0.87890625
                },
                {
                    "start": 2379.73,
                    "end": 2380.61,
                    "word": " multiple",
                    "probability": 0.80029296875
                },
                {
                    "start": 2380.61,
                    "end": 2381.37,
                    "word": " recipients",
                    "probability": 0.6259765625
                },
                {
                    "start": 2381.37,
                    "end": 2381.85,
                    "word": " or",
                    "probability": 0.62890625
                },
                {
                    "start": 2381.85,
                    "end": 2383.21,
                    "word": " carbon",
                    "probability": 0.362548828125
                },
                {
                    "start": 2383.21,
                    "end": 2383.57,
                    "word": " copy",
                    "probability": 0.69921875
                },
                {
                    "start": 2383.57,
                    "end": 2384.29,
                    "word": " recipients",
                    "probability": 0.66015625
                },
                {
                    "start": 2384.29,
                    "end": 2385.09,
                    "word": ".",
                    "probability": 0.9453125
                }
            ]
        },
        {
            "id": 342,
            "seek": 233687,
            "start": 2390.28,
            "end": 2395.1,
            "text": " Next, we will have a look at the performance of our CatBot.",
            "tokens": [
                51276,
                3087,
                11,
                321,
                486,
                362,
                257,
                574,
                412,
                264,
                3389,
                295,
                527,
                9565,
                33,
                310,
                13,
                51452
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2709134615384615,
            "compression_ratio": 1.7430555555555556,
            "no_speech_prob": 0.067138671875,
            "words": [
                {
                    "start": 2390.28,
                    "end": 2391.08,
                    "word": " Next",
                    "probability": 0.837890625
                },
                {
                    "start": 2391.08,
                    "end": 2391.18,
                    "word": ",",
                    "probability": 0.66943359375
                },
                {
                    "start": 2391.18,
                    "end": 2391.34,
                    "word": " we",
                    "probability": 0.904296875
                },
                {
                    "start": 2391.34,
                    "end": 2391.6,
                    "word": " will",
                    "probability": 0.79150390625
                },
                {
                    "start": 2391.6,
                    "end": 2391.78,
                    "word": " have",
                    "probability": 0.84619140625
                },
                {
                    "start": 2391.88,
                    "end": 2392.06,
                    "word": " a look",
                    "probability": 0.92822265625
                },
                {
                    "start": 2392.06,
                    "end": 2392.3,
                    "word": " at",
                    "probability": 0.90771484375
                },
                {
                    "start": 2392.3,
                    "end": 2392.5,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 2392.5,
                    "end": 2392.86,
                    "word": " performance",
                    "probability": 0.833984375
                },
                {
                    "start": 2392.86,
                    "end": 2393.36,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 2393.36,
                    "end": 2393.74,
                    "word": " our",
                    "probability": 0.77099609375
                },
                {
                    "start": 2393.74,
                    "end": 2394.3,
                    "word": " CatBot",
                    "probability": 0.653564453125
                },
                {
                    "start": 2394.3,
                    "end": 2395.1,
                    "word": ".",
                    "probability": 0.94921875
                }
            ]
        },
        {
            "id": 343,
            "seek": 233687,
            "start": 2400.64,
            "end": 2404.66,
            "text": " Let's have a look at the performance of a CatBot.",
            "tokens": [
                51544,
                961,
                311,
                362,
                257,
                574,
                412,
                264,
                3389,
                295,
                257,
                9565,
                33,
                310,
                13,
                51680
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2709134615384615,
            "compression_ratio": 1.7430555555555556,
            "no_speech_prob": 0.067138671875,
            "words": [
                {
                    "start": 2400.64,
                    "end": 2401.44,
                    "word": " Let's",
                    "probability": 0.918701171875
                },
                {
                    "start": 2401.44,
                    "end": 2401.58,
                    "word": " have",
                    "probability": 0.87451171875
                },
                {
                    "start": 2401.68,
                    "end": 2401.82,
                    "word": " a look",
                    "probability": 0.9189453125
                },
                {
                    "start": 2401.82,
                    "end": 2402.0,
                    "word": " at",
                    "probability": 0.90869140625
                },
                {
                    "start": 2402.0,
                    "end": 2402.16,
                    "word": " the",
                    "probability": 0.81689453125
                },
                {
                    "start": 2402.16,
                    "end": 2402.76,
                    "word": " performance",
                    "probability": 0.84130859375
                },
                {
                    "start": 2402.76,
                    "end": 2403.16,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 2403.32,
                    "end": 2403.9,
                    "word": " a CatBot",
                    "probability": 0.9710286458333334
                },
                {
                    "start": 2403.9,
                    "end": 2404.66,
                    "word": ".",
                    "probability": 0.96484375
                }
            ]
        },
        {
            "id": 344,
            "seek": 236397,
            "start": 2406.9,
            "end": 2412.5,
            "text": " We used a data set of 10 million benign samples.",
            "tokens": [
                50412,
                492,
                1143,
                257,
                1412,
                992,
                295,
                1266,
                2459,
                3271,
                788,
                10938,
                13,
                50632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3332386233589866,
            "compression_ratio": 1.4689655172413794,
            "no_speech_prob": 0.006313323974609375,
            "words": [
                {
                    "start": 2406.9,
                    "end": 2407.38,
                    "word": " We",
                    "probability": 0.77392578125
                },
                {
                    "start": 2407.38,
                    "end": 2407.86,
                    "word": " used",
                    "probability": 0.7509765625
                },
                {
                    "start": 2408.12,
                    "end": 2408.5,
                    "word": " a data",
                    "probability": 0.282470703125
                },
                {
                    "start": 2408.5,
                    "end": 2409.1,
                    "word": " set",
                    "probability": 0.8984375
                },
                {
                    "start": 2409.1,
                    "end": 2409.56,
                    "word": " of",
                    "probability": 0.89501953125
                },
                {
                    "start": 2409.56,
                    "end": 2409.94,
                    "word": " 10",
                    "probability": 0.8701171875
                },
                {
                    "start": 2409.94,
                    "end": 2410.3,
                    "word": " million",
                    "probability": 0.59716796875
                },
                {
                    "start": 2410.3,
                    "end": 2410.92,
                    "word": " benign",
                    "probability": 0.7197265625
                },
                {
                    "start": 2410.92,
                    "end": 2411.54,
                    "word": " samples",
                    "probability": 0.6796875
                },
                {
                    "start": 2411.54,
                    "end": 2412.5,
                    "word": ".",
                    "probability": 0.449951171875
                }
            ]
        },
        {
            "id": 345,
            "seek": 236397,
            "start": 2412.72,
            "end": 2422.18,
            "text": " And the data set also includes 350 phishing emails and 1000 BC emails.",
            "tokens": [
                50684,
                400,
                264,
                1412,
                992,
                611,
                5974,
                18065,
                903,
                3807,
                12524,
                293,
                9714,
                14359,
                12524,
                13,
                51116
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3332386233589866,
            "compression_ratio": 1.4689655172413794,
            "no_speech_prob": 0.006313323974609375,
            "words": [
                {
                    "start": 2412.72,
                    "end": 2412.88,
                    "word": " And",
                    "probability": 0.130126953125
                },
                {
                    "start": 2412.88,
                    "end": 2414.04,
                    "word": " the",
                    "probability": 0.765625
                },
                {
                    "start": 2414.04,
                    "end": 2414.32,
                    "word": " data",
                    "probability": 0.671875
                },
                {
                    "start": 2414.32,
                    "end": 2414.68,
                    "word": " set",
                    "probability": 0.85205078125
                },
                {
                    "start": 2414.68,
                    "end": 2415.0,
                    "word": " also",
                    "probability": 0.7392578125
                },
                {
                    "start": 2415.0,
                    "end": 2415.52,
                    "word": " includes",
                    "probability": 0.410400390625
                },
                {
                    "start": 2415.52,
                    "end": 2417.36,
                    "word": " 350",
                    "probability": 0.67529296875
                },
                {
                    "start": 2417.36,
                    "end": 2418.36,
                    "word": " phishing",
                    "probability": 0.7890625
                },
                {
                    "start": 2418.36,
                    "end": 2418.72,
                    "word": " emails",
                    "probability": 0.69189453125
                },
                {
                    "start": 2418.72,
                    "end": 2419.2,
                    "word": " and",
                    "probability": 0.83447265625
                },
                {
                    "start": 2419.2,
                    "end": 2419.94,
                    "word": " 1000",
                    "probability": 0.3447265625
                },
                {
                    "start": 2419.94,
                    "end": 2420.5,
                    "word": " BC",
                    "probability": 0.232666015625
                },
                {
                    "start": 2420.5,
                    "end": 2421.22,
                    "word": " emails",
                    "probability": 0.689453125
                },
                {
                    "start": 2421.22,
                    "end": 2422.18,
                    "word": ".",
                    "probability": 0.9677734375
                }
            ]
        },
        {
            "id": 346,
            "seek": 236397,
            "start": 2423.74,
            "end": 2433.22,
            "text": " We used time split to allocate 70% of samples as training and remaining 30% for test samples.",
            "tokens": [
                51252,
                492,
                1143,
                565,
                7472,
                281,
                35713,
                5285,
                4,
                295,
                10938,
                382,
                3097,
                293,
                8877,
                2217,
                4,
                337,
                1500,
                10938,
                13,
                51696
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3332386233589866,
            "compression_ratio": 1.4689655172413794,
            "no_speech_prob": 0.006313323974609375,
            "words": [
                {
                    "start": 2423.74,
                    "end": 2424.22,
                    "word": " We",
                    "probability": 0.90185546875
                },
                {
                    "start": 2424.22,
                    "end": 2424.52,
                    "word": " used",
                    "probability": 0.56689453125
                },
                {
                    "start": 2424.52,
                    "end": 2424.9,
                    "word": " time",
                    "probability": 0.35107421875
                },
                {
                    "start": 2424.9,
                    "end": 2425.28,
                    "word": " split",
                    "probability": 0.7158203125
                },
                {
                    "start": 2425.28,
                    "end": 2425.62,
                    "word": " to",
                    "probability": 0.869140625
                },
                {
                    "start": 2425.62,
                    "end": 2426.08,
                    "word": " allocate",
                    "probability": 0.445068359375
                },
                {
                    "start": 2426.08,
                    "end": 2426.64,
                    "word": " 70",
                    "probability": 0.892578125
                },
                {
                    "start": 2426.64,
                    "end": 2427.12,
                    "word": "%",
                    "probability": 0.8857421875
                },
                {
                    "start": 2427.12,
                    "end": 2427.44,
                    "word": " of",
                    "probability": 0.87744140625
                },
                {
                    "start": 2427.44,
                    "end": 2428.48,
                    "word": " samples",
                    "probability": 0.62548828125
                },
                {
                    "start": 2428.48,
                    "end": 2428.74,
                    "word": " as",
                    "probability": 0.89404296875
                },
                {
                    "start": 2428.74,
                    "end": 2429.22,
                    "word": " training",
                    "probability": 0.7822265625
                },
                {
                    "start": 2429.22,
                    "end": 2429.62,
                    "word": " and",
                    "probability": 0.81298828125
                },
                {
                    "start": 2429.62,
                    "end": 2430.44,
                    "word": " remaining",
                    "probability": 0.7412109375
                },
                {
                    "start": 2430.44,
                    "end": 2430.92,
                    "word": " 30",
                    "probability": 0.8837890625
                },
                {
                    "start": 2430.92,
                    "end": 2431.5,
                    "word": "%",
                    "probability": 0.994140625
                },
                {
                    "start": 2431.5,
                    "end": 2431.86,
                    "word": " for",
                    "probability": 0.88720703125
                },
                {
                    "start": 2431.86,
                    "end": 2432.26,
                    "word": " test",
                    "probability": 0.7978515625
                },
                {
                    "start": 2432.26,
                    "end": 2432.74,
                    "word": " samples",
                    "probability": 0.75927734375
                },
                {
                    "start": 2432.74,
                    "end": 2433.22,
                    "word": ".",
                    "probability": 0.99560546875
                }
            ]
        },
        {
            "id": 347,
            "seek": 238403,
            "start": 2434.01,
            "end": 2444.14,
            "text": " We included two baseline models to compare CatBot. The first one is DistilBot, which has",
            "tokens": [
                50408,
                492,
                5556,
                732,
                20518,
                5245,
                281,
                6794,
                9565,
                33,
                310,
                13,
                440,
                700,
                472,
                307,
                9840,
                388,
                33,
                310,
                11,
                597,
                575,
                50908
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28265623569488524,
            "compression_ratio": 1.2986111111111112,
            "no_speech_prob": 0.009429931640625,
            "words": [
                {
                    "start": 2434.01,
                    "end": 2434.46,
                    "word": " We",
                    "probability": 0.77001953125
                },
                {
                    "start": 2434.46,
                    "end": 2436.5,
                    "word": " included",
                    "probability": 0.74267578125
                },
                {
                    "start": 2436.5,
                    "end": 2437.14,
                    "word": " two",
                    "probability": 0.7548828125
                },
                {
                    "start": 2437.14,
                    "end": 2437.84,
                    "word": " baseline",
                    "probability": 0.685546875
                },
                {
                    "start": 2437.84,
                    "end": 2438.44,
                    "word": " models",
                    "probability": 0.859375
                },
                {
                    "start": 2438.44,
                    "end": 2438.66,
                    "word": " to",
                    "probability": 0.2435302734375
                },
                {
                    "start": 2438.66,
                    "end": 2439.02,
                    "word": " compare",
                    "probability": 0.8681640625
                },
                {
                    "start": 2439.02,
                    "end": 2439.88,
                    "word": " CatBot",
                    "probability": 0.6328531901041666
                },
                {
                    "start": 2439.88,
                    "end": 2440.78,
                    "word": ".",
                    "probability": 0.81005859375
                },
                {
                    "start": 2441.02,
                    "end": 2441.24,
                    "word": " The",
                    "probability": 0.7783203125
                },
                {
                    "start": 2441.24,
                    "end": 2441.58,
                    "word": " first",
                    "probability": 0.7958984375
                },
                {
                    "start": 2441.58,
                    "end": 2442.06,
                    "word": " one",
                    "probability": 0.7509765625
                },
                {
                    "start": 2442.06,
                    "end": 2442.64,
                    "word": " is",
                    "probability": 0.8974609375
                },
                {
                    "start": 2442.64,
                    "end": 2443.44,
                    "word": " DistilBot",
                    "probability": 0.7601318359375
                },
                {
                    "start": 2443.44,
                    "end": 2443.6,
                    "word": ",",
                    "probability": 0.5498046875
                },
                {
                    "start": 2443.6,
                    "end": 2443.86,
                    "word": " which",
                    "probability": 0.919921875
                },
                {
                    "start": 2443.86,
                    "end": 2444.14,
                    "word": " has",
                    "probability": 0.8935546875
                }
            ]
        },
        {
            "id": 348,
            "seek": 238403,
            "start": 2444.14,
            "end": 2453.28,
            "text": " six transformer blocks. And another one is LSTM Long Short Memory. The model is a recurrent neural",
            "tokens": [
                50908,
                2309,
                31782,
                8474,
                13,
                400,
                1071,
                472,
                307,
                441,
                6840,
                44,
                8282,
                16881,
                38203,
                13,
                440,
                2316,
                307,
                257,
                18680,
                1753,
                18161,
                51364
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28265623569488524,
            "compression_ratio": 1.2986111111111112,
            "no_speech_prob": 0.009429931640625,
            "words": [
                {
                    "start": 2444.14,
                    "end": 2444.46,
                    "word": " six",
                    "probability": 0.4912109375
                },
                {
                    "start": 2444.46,
                    "end": 2445.04,
                    "word": " transformer",
                    "probability": 0.7451171875
                },
                {
                    "start": 2445.04,
                    "end": 2445.52,
                    "word": " blocks",
                    "probability": 0.80419921875
                },
                {
                    "start": 2445.52,
                    "end": 2446.42,
                    "word": ".",
                    "probability": 0.6591796875
                },
                {
                    "start": 2446.76,
                    "end": 2447.04,
                    "word": " And",
                    "probability": 0.65380859375
                },
                {
                    "start": 2447.04,
                    "end": 2447.46,
                    "word": " another",
                    "probability": 0.7890625
                },
                {
                    "start": 2447.46,
                    "end": 2447.78,
                    "word": " one",
                    "probability": 0.7900390625
                },
                {
                    "start": 2447.78,
                    "end": 2447.96,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 2447.96,
                    "end": 2448.62,
                    "word": " LSTM",
                    "probability": 0.8351236979166666
                },
                {
                    "start": 2448.62,
                    "end": 2448.96,
                    "word": " Long",
                    "probability": 0.276611328125
                },
                {
                    "start": 2448.96,
                    "end": 2449.3,
                    "word": " Short",
                    "probability": 0.56298828125
                },
                {
                    "start": 2449.3,
                    "end": 2449.8,
                    "word": " Memory",
                    "probability": 0.85400390625
                },
                {
                    "start": 2449.8,
                    "end": 2450.66,
                    "word": ".",
                    "probability": 0.92626953125
                },
                {
                    "start": 2450.66,
                    "end": 2450.94,
                    "word": " The",
                    "probability": 0.79541015625
                },
                {
                    "start": 2450.94,
                    "end": 2451.36,
                    "word": " model",
                    "probability": 0.82470703125
                },
                {
                    "start": 2451.36,
                    "end": 2451.92,
                    "word": " is",
                    "probability": 0.8759765625
                },
                {
                    "start": 2452.28,
                    "end": 2452.9,
                    "word": " a recurrent",
                    "probability": 0.861328125
                },
                {
                    "start": 2452.9,
                    "end": 2453.28,
                    "word": " neural",
                    "probability": 0.916015625
                }
            ]
        },
        {
            "id": 349,
            "seek": 241013,
            "start": 2453.28,
            "end": 2465.14,
            "text": " network architecture, which also uses BERT's same embedding layer. We trained three models on a",
            "tokens": [
                50364,
                3209,
                9482,
                11,
                597,
                611,
                4960,
                363,
                31479,
                311,
                912,
                12240,
                3584,
                4583,
                13,
                492,
                8895,
                1045,
                5245,
                322,
                257,
                50864
            ],
            "temperature": 0.0,
            "avg_logprob": -0.36136363311247394,
            "compression_ratio": 1.3657142857142857,
            "no_speech_prob": 0.2298583984375,
            "words": [
                {
                    "start": 2453.28,
                    "end": 2453.72,
                    "word": " network",
                    "probability": 0.469970703125
                },
                {
                    "start": 2453.72,
                    "end": 2454.52,
                    "word": " architecture",
                    "probability": 0.86279296875
                },
                {
                    "start": 2454.52,
                    "end": 2455.9,
                    "word": ",",
                    "probability": 0.57666015625
                },
                {
                    "start": 2455.9,
                    "end": 2456.24,
                    "word": " which",
                    "probability": 0.91650390625
                },
                {
                    "start": 2456.24,
                    "end": 2456.54,
                    "word": " also",
                    "probability": 0.8125
                },
                {
                    "start": 2456.54,
                    "end": 2456.98,
                    "word": " uses",
                    "probability": 0.04443359375
                },
                {
                    "start": 2456.98,
                    "end": 2457.9,
                    "word": " BERT's",
                    "probability": 0.7317708333333334
                },
                {
                    "start": 2457.9,
                    "end": 2458.46,
                    "word": " same",
                    "probability": 0.75244140625
                },
                {
                    "start": 2458.46,
                    "end": 2459.0,
                    "word": " embedding",
                    "probability": 0.88330078125
                },
                {
                    "start": 2459.0,
                    "end": 2459.68,
                    "word": " layer",
                    "probability": 0.7412109375
                },
                {
                    "start": 2459.68,
                    "end": 2460.76,
                    "word": ".",
                    "probability": 0.96630859375
                },
                {
                    "start": 2460.76,
                    "end": 2461.04,
                    "word": " We",
                    "probability": 0.89208984375
                },
                {
                    "start": 2461.04,
                    "end": 2461.42,
                    "word": " trained",
                    "probability": 0.79052734375
                },
                {
                    "start": 2461.42,
                    "end": 2461.94,
                    "word": " three",
                    "probability": 0.6044921875
                },
                {
                    "start": 2461.94,
                    "end": 2462.82,
                    "word": " models",
                    "probability": 0.82080078125
                },
                {
                    "start": 2462.82,
                    "end": 2463.18,
                    "word": " on",
                    "probability": 0.90087890625
                },
                {
                    "start": 2464.68,
                    "end": 2465.14,
                    "word": " a GPU",
                    "probability": 0.8115234375
                }
            ]
        },
        {
            "id": 350,
            "seek": 241013,
            "start": 2465.14,
            "end": 2472.92,
            "text": " GPU instance from AWS. And we assigned high sample weight for the",
            "tokens": [
                50936,
                18407,
                5197,
                490,
                17650,
                13,
                400,
                321,
                13279,
                1090,
                6889,
                3364,
                337,
                264,
                51256
            ],
            "temperature": 0.0,
            "avg_logprob": -0.36136363311247394,
            "compression_ratio": 1.3657142857142857,
            "no_speech_prob": 0.2298583984375,
            "words": [
                {
                    "start": 2465.14,
                    "end": 2465.72,
                    "word": " instance",
                    "probability": 0.85791015625
                },
                {
                    "start": 2465.72,
                    "end": 2466.08,
                    "word": " from",
                    "probability": 0.7802734375
                },
                {
                    "start": 2466.08,
                    "end": 2466.78,
                    "word": " AWS",
                    "probability": 0.88916015625
                },
                {
                    "start": 2466.78,
                    "end": 2467.9,
                    "word": ".",
                    "probability": 0.33447265625
                },
                {
                    "start": 2468.04,
                    "end": 2468.7,
                    "word": " And",
                    "probability": 0.67724609375
                },
                {
                    "start": 2468.7,
                    "end": 2468.92,
                    "word": " we",
                    "probability": 0.83984375
                },
                {
                    "start": 2468.92,
                    "end": 2469.38,
                    "word": " assigned",
                    "probability": 0.78564453125
                },
                {
                    "start": 2469.38,
                    "end": 2469.72,
                    "word": " high",
                    "probability": 0.7548828125
                },
                {
                    "start": 2469.72,
                    "end": 2470.2,
                    "word": " sample",
                    "probability": 0.73681640625
                },
                {
                    "start": 2470.2,
                    "end": 2470.58,
                    "word": " weight",
                    "probability": 0.5595703125
                },
                {
                    "start": 2470.58,
                    "end": 2471.0,
                    "word": " for",
                    "probability": 0.8974609375
                },
                {
                    "start": 2471.0,
                    "end": 2471.82,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 2471.82,
                    "end": 2472.92,
                    "word": " PEC",
                    "probability": 0.6787109375
                }
            ]
        },
        {
            "id": 351,
            "seek": 241013,
            "start": 2472.92,
            "end": 2479.38,
            "text": " PEC samples and oversample minor class malicious samples to allocate balanced",
            "tokens": [
                51320,
                430,
                8140,
                10938,
                293,
                15488,
                335,
                781,
                6696,
                1508,
                33496,
                10938,
                281,
                35713,
                13902,
                51664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.36136363311247394,
            "compression_ratio": 1.3657142857142857,
            "no_speech_prob": 0.2298583984375,
            "words": [
                {
                    "start": 2472.92,
                    "end": 2473.62,
                    "word": " samples",
                    "probability": 0.7646484375
                },
                {
                    "start": 2473.62,
                    "end": 2474.26,
                    "word": " and",
                    "probability": 0.658203125
                },
                {
                    "start": 2474.26,
                    "end": 2475.08,
                    "word": " oversample",
                    "probability": 0.6907552083333334
                },
                {
                    "start": 2475.08,
                    "end": 2476.08,
                    "word": " minor",
                    "probability": 0.3037109375
                },
                {
                    "start": 2476.08,
                    "end": 2476.78,
                    "word": " class",
                    "probability": 0.73095703125
                },
                {
                    "start": 2476.78,
                    "end": 2477.38,
                    "word": " malicious",
                    "probability": 0.1348876953125
                },
                {
                    "start": 2477.38,
                    "end": 2477.82,
                    "word": " samples",
                    "probability": 0.74951171875
                },
                {
                    "start": 2477.82,
                    "end": 2478.1,
                    "word": " to",
                    "probability": 0.8974609375
                },
                {
                    "start": 2478.1,
                    "end": 2478.64,
                    "word": " allocate",
                    "probability": 0.63232421875
                },
                {
                    "start": 2478.64,
                    "end": 2479.38,
                    "word": " balanced",
                    "probability": 0.81640625
                }
            ]
        },
        {
            "id": 352,
            "seek": 243943,
            "start": 2480.12,
            "end": 2489.24,
            "text": " samples in each mini-matching. To compare performance, we use R-Curves and L-Re-Under-Curve.",
            "tokens": [
                50404,
                10938,
                294,
                1184,
                8382,
                12,
                15677,
                66,
                571,
                13,
                1407,
                6794,
                3389,
                11,
                321,
                764,
                497,
                12,
                34,
                374,
                977,
                293,
                441,
                12,
                8524,
                12,
                52,
                20535,
                12,
                34,
                374,
                303,
                13,
                50848
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30979939448980637,
            "compression_ratio": 1.5647058823529412,
            "no_speech_prob": 0.0088958740234375,
            "words": [
                {
                    "start": 2480.12,
                    "end": 2480.6,
                    "word": " samples",
                    "probability": 0.0323486328125
                },
                {
                    "start": 2480.6,
                    "end": 2481.06,
                    "word": " in",
                    "probability": 0.8359375
                },
                {
                    "start": 2481.06,
                    "end": 2481.52,
                    "word": " each",
                    "probability": 0.78759765625
                },
                {
                    "start": 2481.52,
                    "end": 2481.76,
                    "word": " mini",
                    "probability": 0.720703125
                },
                {
                    "start": 2481.76,
                    "end": 2482.2,
                    "word": "-matching",
                    "probability": 0.67474365234375
                },
                {
                    "start": 2482.2,
                    "end": 2483.16,
                    "word": ".",
                    "probability": 0.859375
                },
                {
                    "start": 2483.68,
                    "end": 2483.94,
                    "word": " To",
                    "probability": 0.88134765625
                },
                {
                    "start": 2483.94,
                    "end": 2484.3,
                    "word": " compare",
                    "probability": 0.8603515625
                },
                {
                    "start": 2484.3,
                    "end": 2484.84,
                    "word": " performance",
                    "probability": 0.7724609375
                },
                {
                    "start": 2484.84,
                    "end": 2485.04,
                    "word": ",",
                    "probability": 0.80224609375
                },
                {
                    "start": 2485.04,
                    "end": 2485.22,
                    "word": " we",
                    "probability": 0.8916015625
                },
                {
                    "start": 2485.22,
                    "end": 2486.42,
                    "word": " use",
                    "probability": 0.6796875
                },
                {
                    "start": 2486.42,
                    "end": 2486.7,
                    "word": " R",
                    "probability": 0.37060546875
                },
                {
                    "start": 2486.7,
                    "end": 2487.26,
                    "word": "-Curves",
                    "probability": 0.57122802734375
                },
                {
                    "start": 2487.26,
                    "end": 2487.9,
                    "word": " and",
                    "probability": 0.8681640625
                },
                {
                    "start": 2487.9,
                    "end": 2488.2,
                    "word": " L",
                    "probability": 0.30810546875
                },
                {
                    "start": 2488.2,
                    "end": 2488.34,
                    "word": "-Re",
                    "probability": 0.5548095703125
                },
                {
                    "start": 2488.34,
                    "end": 2488.56,
                    "word": "-Under",
                    "probability": 0.6275634765625
                },
                {
                    "start": 2488.56,
                    "end": 2488.96,
                    "word": "-Curve",
                    "probability": 0.8138427734375
                },
                {
                    "start": 2488.96,
                    "end": 2489.24,
                    "word": ".",
                    "probability": 0.7685546875
                }
            ]
        },
        {
            "id": 353,
            "seek": 243943,
            "start": 2489.24,
            "end": 2496.24,
            "text": " Also, we compare inference speed and the model size as key performance metrics.",
            "tokens": [
                50848,
                2743,
                11,
                321,
                6794,
                38253,
                3073,
                293,
                264,
                2316,
                2744,
                382,
                2141,
                3389,
                16367,
                13,
                51156
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30979939448980637,
            "compression_ratio": 1.5647058823529412,
            "no_speech_prob": 0.0088958740234375,
            "words": [
                {
                    "start": 2489.24,
                    "end": 2489.48,
                    "word": " Also",
                    "probability": 0.88330078125
                },
                {
                    "start": 2489.48,
                    "end": 2489.7,
                    "word": ",",
                    "probability": 0.6787109375
                },
                {
                    "start": 2489.7,
                    "end": 2489.7,
                    "word": " we",
                    "probability": 0.8935546875
                },
                {
                    "start": 2489.7,
                    "end": 2490.2,
                    "word": " compare",
                    "probability": 0.83349609375
                },
                {
                    "start": 2490.2,
                    "end": 2491.28,
                    "word": " inference",
                    "probability": 0.591796875
                },
                {
                    "start": 2491.28,
                    "end": 2491.7,
                    "word": " speed",
                    "probability": 0.806640625
                },
                {
                    "start": 2491.7,
                    "end": 2492.02,
                    "word": " and",
                    "probability": 0.88671875
                },
                {
                    "start": 2492.02,
                    "end": 2492.42,
                    "word": " the",
                    "probability": 0.183837890625
                },
                {
                    "start": 2492.42,
                    "end": 2492.7,
                    "word": " model",
                    "probability": 0.7734375
                },
                {
                    "start": 2492.7,
                    "end": 2493.16,
                    "word": " size",
                    "probability": 0.80615234375
                },
                {
                    "start": 2493.16,
                    "end": 2493.64,
                    "word": " as",
                    "probability": 0.8974609375
                },
                {
                    "start": 2493.64,
                    "end": 2494.12,
                    "word": " key",
                    "probability": 0.76708984375
                },
                {
                    "start": 2494.12,
                    "end": 2494.7,
                    "word": " performance",
                    "probability": 0.78369140625
                },
                {
                    "start": 2494.7,
                    "end": 2495.28,
                    "word": " metrics",
                    "probability": 0.75830078125
                },
                {
                    "start": 2495.28,
                    "end": 2496.24,
                    "word": ".",
                    "probability": 0.97314453125
                }
            ]
        },
        {
            "id": 354,
            "seek": 243943,
            "start": 2500.28,
            "end": 2511.78,
            "text": " These R-Curves compare our CatBot model with two baseline models. The top blue one is CatBot,",
            "tokens": [
                51248,
                1981,
                497,
                12,
                34,
                374,
                977,
                6794,
                527,
                9565,
                33,
                310,
                2316,
                365,
                732,
                20518,
                5245,
                13,
                440,
                1192,
                3344,
                472,
                307,
                9565,
                33,
                310,
                11,
                51828
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30979939448980637,
            "compression_ratio": 1.5647058823529412,
            "no_speech_prob": 0.0088958740234375,
            "words": [
                {
                    "start": 2500.28,
                    "end": 2500.54,
                    "word": " These",
                    "probability": 0.80126953125
                },
                {
                    "start": 2500.54,
                    "end": 2500.88,
                    "word": " R",
                    "probability": 0.9697265625
                },
                {
                    "start": 2500.88,
                    "end": 2501.4,
                    "word": "-Curves",
                    "probability": 0.920166015625
                },
                {
                    "start": 2501.4,
                    "end": 2502.02,
                    "word": " compare",
                    "probability": 0.83447265625
                },
                {
                    "start": 2502.02,
                    "end": 2502.64,
                    "word": " our",
                    "probability": 0.80615234375
                },
                {
                    "start": 2502.64,
                    "end": 2503.3,
                    "word": " CatBot",
                    "probability": 0.5765787760416666
                },
                {
                    "start": 2503.3,
                    "end": 2504.22,
                    "word": " model",
                    "probability": 0.75
                },
                {
                    "start": 2504.22,
                    "end": 2506.12,
                    "word": " with",
                    "probability": 0.794921875
                },
                {
                    "start": 2506.12,
                    "end": 2506.92,
                    "word": " two",
                    "probability": 0.755859375
                },
                {
                    "start": 2506.92,
                    "end": 2507.42,
                    "word": " baseline",
                    "probability": 0.6982421875
                },
                {
                    "start": 2507.42,
                    "end": 2508.12,
                    "word": " models",
                    "probability": 0.84619140625
                },
                {
                    "start": 2508.12,
                    "end": 2508.94,
                    "word": ".",
                    "probability": 0.96337890625
                },
                {
                    "start": 2508.94,
                    "end": 2509.42,
                    "word": " The",
                    "probability": 0.8349609375
                },
                {
                    "start": 2509.42,
                    "end": 2509.92,
                    "word": " top",
                    "probability": 0.8291015625
                },
                {
                    "start": 2509.92,
                    "end": 2510.46,
                    "word": " blue",
                    "probability": 0.8505859375
                },
                {
                    "start": 2510.46,
                    "end": 2510.72,
                    "word": " one",
                    "probability": 0.80810546875
                },
                {
                    "start": 2510.72,
                    "end": 2510.92,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 2510.92,
                    "end": 2511.56,
                    "word": " CatBot",
                    "probability": 0.9773763020833334
                },
                {
                    "start": 2511.56,
                    "end": 2511.78,
                    "word": ",",
                    "probability": 0.374267578125
                }
            ]
        },
        {
            "id": 355,
            "seek": 246225,
            "start": 2511.78,
            "end": 2518.14,
            "text": " and the second one is DistilBot, and the bottom green one is LSTM.",
            "tokens": [
                50364,
                293,
                264,
                1150,
                472,
                307,
                9840,
                388,
                33,
                310,
                11,
                293,
                264,
                2767,
                3092,
                472,
                307,
                441,
                6840,
                44,
                13,
                50628
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23327850458914773,
            "compression_ratio": 1.3777777777777778,
            "no_speech_prob": 0.0005469322204589844,
            "words": [
                {
                    "start": 2511.78,
                    "end": 2512.06,
                    "word": " and",
                    "probability": 0.328857421875
                },
                {
                    "start": 2512.06,
                    "end": 2512.3,
                    "word": " the",
                    "probability": 0.1868896484375
                },
                {
                    "start": 2512.3,
                    "end": 2512.52,
                    "word": " second",
                    "probability": 0.6884765625
                },
                {
                    "start": 2512.52,
                    "end": 2513.14,
                    "word": " one",
                    "probability": 0.7294921875
                },
                {
                    "start": 2513.14,
                    "end": 2513.68,
                    "word": " is",
                    "probability": 0.89208984375
                },
                {
                    "start": 2513.68,
                    "end": 2514.42,
                    "word": " DistilBot",
                    "probability": 0.466552734375
                },
                {
                    "start": 2514.42,
                    "end": 2514.54,
                    "word": ",",
                    "probability": 0.272705078125
                },
                {
                    "start": 2514.54,
                    "end": 2514.68,
                    "word": " and",
                    "probability": 0.85205078125
                },
                {
                    "start": 2514.68,
                    "end": 2514.8,
                    "word": " the",
                    "probability": 0.79443359375
                },
                {
                    "start": 2514.8,
                    "end": 2515.16,
                    "word": " bottom",
                    "probability": 0.79296875
                },
                {
                    "start": 2515.16,
                    "end": 2515.72,
                    "word": " green",
                    "probability": 0.72265625
                },
                {
                    "start": 2515.72,
                    "end": 2516.1,
                    "word": " one",
                    "probability": 0.80908203125
                },
                {
                    "start": 2516.1,
                    "end": 2516.34,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 2516.34,
                    "end": 2517.14,
                    "word": " LSTM",
                    "probability": 0.9480794270833334
                },
                {
                    "start": 2517.14,
                    "end": 2518.14,
                    "word": ".",
                    "probability": 0.91162109375
                }
            ]
        },
        {
            "id": 356,
            "seek": 246225,
            "start": 2518.38,
            "end": 2523.14,
            "text": " Our model outperformed the two baseline models.",
            "tokens": [
                50700,
                2621,
                2316,
                484,
                610,
                22892,
                264,
                732,
                20518,
                5245,
                13,
                50880
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23327850458914773,
            "compression_ratio": 1.3777777777777778,
            "no_speech_prob": 0.0005469322204589844,
            "words": [
                {
                    "start": 2518.38,
                    "end": 2518.8,
                    "word": " Our",
                    "probability": 0.8818359375
                },
                {
                    "start": 2518.8,
                    "end": 2519.24,
                    "word": " model",
                    "probability": 0.67626953125
                },
                {
                    "start": 2519.24,
                    "end": 2519.94,
                    "word": " outperformed",
                    "probability": 0.7760416666666666
                },
                {
                    "start": 2519.94,
                    "end": 2520.2,
                    "word": " the",
                    "probability": 0.62451171875
                },
                {
                    "start": 2520.2,
                    "end": 2520.44,
                    "word": " two",
                    "probability": 0.7802734375
                },
                {
                    "start": 2520.44,
                    "end": 2521.02,
                    "word": " baseline",
                    "probability": 0.861328125
                },
                {
                    "start": 2521.02,
                    "end": 2522.14,
                    "word": " models",
                    "probability": 0.794921875
                },
                {
                    "start": 2522.14,
                    "end": 2523.14,
                    "word": ".",
                    "probability": 0.56201171875
                }
            ]
        },
        {
            "id": 357,
            "seek": 246225,
            "start": 2524.34,
            "end": 2534.6,
            "text": " And our model achieved 0.82% positive rate at 0.1% false positive rate.",
            "tokens": [
                50992,
                400,
                527,
                2316,
                11042,
                1958,
                13,
                32848,
                4,
                3353,
                3314,
                412,
                1958,
                13,
                16,
                4,
                7908,
                3353,
                3314,
                13,
                51468
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23327850458914773,
            "compression_ratio": 1.3777777777777778,
            "no_speech_prob": 0.0005469322204589844,
            "words": [
                {
                    "start": 2524.34,
                    "end": 2524.62,
                    "word": " And",
                    "probability": 0.444091796875
                },
                {
                    "start": 2524.62,
                    "end": 2526.08,
                    "word": " our",
                    "probability": 0.7744140625
                },
                {
                    "start": 2526.08,
                    "end": 2526.5,
                    "word": " model",
                    "probability": 0.787109375
                },
                {
                    "start": 2526.5,
                    "end": 2527.74,
                    "word": " achieved",
                    "probability": 0.69580078125
                },
                {
                    "start": 2527.74,
                    "end": 2528.32,
                    "word": " 0",
                    "probability": 0.875
                },
                {
                    "start": 2528.32,
                    "end": 2528.82,
                    "word": ".82",
                    "probability": 0.852783203125
                },
                {
                    "start": 2528.82,
                    "end": 2528.82,
                    "word": "%",
                    "probability": 0.320068359375
                },
                {
                    "start": 2528.82,
                    "end": 2529.62,
                    "word": " positive",
                    "probability": 0.74462890625
                },
                {
                    "start": 2529.62,
                    "end": 2530.2,
                    "word": " rate",
                    "probability": 0.87939453125
                },
                {
                    "start": 2530.2,
                    "end": 2530.62,
                    "word": " at",
                    "probability": 0.68505859375
                },
                {
                    "start": 2530.62,
                    "end": 2530.96,
                    "word": " 0",
                    "probability": 0.95849609375
                },
                {
                    "start": 2530.96,
                    "end": 2531.42,
                    "word": ".1",
                    "probability": 0.96337890625
                },
                {
                    "start": 2531.42,
                    "end": 2532.02,
                    "word": "%",
                    "probability": 0.958984375
                },
                {
                    "start": 2532.02,
                    "end": 2532.36,
                    "word": " false",
                    "probability": 0.291748046875
                },
                {
                    "start": 2532.36,
                    "end": 2532.96,
                    "word": " positive",
                    "probability": 0.84326171875
                },
                {
                    "start": 2532.96,
                    "end": 2533.92,
                    "word": " rate",
                    "probability": 0.873046875
                },
                {
                    "start": 2533.92,
                    "end": 2534.6,
                    "word": ".",
                    "probability": 0.96923828125
                }
            ]
        },
        {
            "id": 358,
            "seek": 248351,
            "start": 2537.41,
            "end": 2547.61,
            "text": " Next, we compare the performance when we removed adapters and context input rated layers.",
            "tokens": [
                50412,
                3087,
                11,
                321,
                6794,
                264,
                3389,
                562,
                321,
                7261,
                23169,
                1559,
                293,
                4319,
                4846,
                22103,
                7914,
                13,
                50884
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2719414792162307,
            "compression_ratio": 1.5,
            "no_speech_prob": 0.0021038055419921875,
            "words": [
                {
                    "start": 2537.41,
                    "end": 2537.93,
                    "word": " Next",
                    "probability": 0.751953125
                },
                {
                    "start": 2537.93,
                    "end": 2538.39,
                    "word": ",",
                    "probability": 0.7529296875
                },
                {
                    "start": 2538.39,
                    "end": 2538.87,
                    "word": " we",
                    "probability": 0.86181640625
                },
                {
                    "start": 2538.87,
                    "end": 2539.49,
                    "word": " compare",
                    "probability": 0.7548828125
                },
                {
                    "start": 2539.49,
                    "end": 2540.13,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 2540.13,
                    "end": 2540.75,
                    "word": " performance",
                    "probability": 0.7958984375
                },
                {
                    "start": 2540.75,
                    "end": 2541.13,
                    "word": " when",
                    "probability": 0.87353515625
                },
                {
                    "start": 2541.13,
                    "end": 2541.47,
                    "word": " we",
                    "probability": 0.875
                },
                {
                    "start": 2541.47,
                    "end": 2542.17,
                    "word": " removed",
                    "probability": 0.52978515625
                },
                {
                    "start": 2542.17,
                    "end": 2543.65,
                    "word": " adapters",
                    "probability": 0.6925048828125
                },
                {
                    "start": 2543.65,
                    "end": 2544.01,
                    "word": " and",
                    "probability": 0.87744140625
                },
                {
                    "start": 2544.01,
                    "end": 2544.65,
                    "word": " context",
                    "probability": 0.68603515625
                },
                {
                    "start": 2544.65,
                    "end": 2545.09,
                    "word": " input",
                    "probability": 0.73681640625
                },
                {
                    "start": 2545.09,
                    "end": 2545.77,
                    "word": " rated",
                    "probability": 0.40673828125
                },
                {
                    "start": 2545.77,
                    "end": 2547.01,
                    "word": " layers",
                    "probability": 0.8076171875
                },
                {
                    "start": 2547.01,
                    "end": 2547.61,
                    "word": ".",
                    "probability": 0.9580078125
                }
            ]
        },
        {
            "id": 359,
            "seek": 248351,
            "start": 2548.11,
            "end": 2557.71,
            "text": " The top one is CatBot, and the second orange one is when we removed adapters. And the bottom one",
            "tokens": [
                50952,
                440,
                1192,
                472,
                307,
                9565,
                33,
                310,
                11,
                293,
                264,
                1150,
                7671,
                472,
                307,
                562,
                321,
                7261,
                23169,
                1559,
                13,
                400,
                264,
                2767,
                472,
                51412
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2719414792162307,
            "compression_ratio": 1.5,
            "no_speech_prob": 0.0021038055419921875,
            "words": [
                {
                    "start": 2548.11,
                    "end": 2548.59,
                    "word": " The",
                    "probability": 0.8076171875
                },
                {
                    "start": 2548.59,
                    "end": 2548.95,
                    "word": " top",
                    "probability": 0.86962890625
                },
                {
                    "start": 2548.95,
                    "end": 2549.27,
                    "word": " one",
                    "probability": 0.7978515625
                },
                {
                    "start": 2549.27,
                    "end": 2549.73,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 2549.73,
                    "end": 2550.37,
                    "word": " CatBot",
                    "probability": 0.6734212239583334
                },
                {
                    "start": 2550.37,
                    "end": 2550.83,
                    "word": ",",
                    "probability": 0.457763671875
                },
                {
                    "start": 2550.83,
                    "end": 2551.27,
                    "word": " and",
                    "probability": 0.7109375
                },
                {
                    "start": 2551.27,
                    "end": 2551.49,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 2551.49,
                    "end": 2551.79,
                    "word": " second",
                    "probability": 0.81689453125
                },
                {
                    "start": 2551.79,
                    "end": 2552.27,
                    "word": " orange",
                    "probability": 0.8486328125
                },
                {
                    "start": 2552.27,
                    "end": 2552.69,
                    "word": " one",
                    "probability": 0.7958984375
                },
                {
                    "start": 2552.69,
                    "end": 2553.21,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 2553.21,
                    "end": 2553.75,
                    "word": " when",
                    "probability": 0.85205078125
                },
                {
                    "start": 2553.75,
                    "end": 2554.29,
                    "word": " we",
                    "probability": 0.89404296875
                },
                {
                    "start": 2554.29,
                    "end": 2554.69,
                    "word": " removed",
                    "probability": 0.7646484375
                },
                {
                    "start": 2554.69,
                    "end": 2555.45,
                    "word": " adapters",
                    "probability": 0.82763671875
                },
                {
                    "start": 2555.45,
                    "end": 2556.03,
                    "word": ".",
                    "probability": 0.3759765625
                },
                {
                    "start": 2556.03,
                    "end": 2556.43,
                    "word": " And",
                    "probability": 0.53759765625
                },
                {
                    "start": 2556.43,
                    "end": 2556.97,
                    "word": " the",
                    "probability": 0.77294921875
                },
                {
                    "start": 2556.97,
                    "end": 2557.27,
                    "word": " bottom",
                    "probability": 0.77001953125
                },
                {
                    "start": 2557.27,
                    "end": 2557.71,
                    "word": " one",
                    "probability": 0.79541015625
                }
            ]
        },
        {
            "id": 360,
            "seek": 250979,
            "start": 2558.06,
            "end": 2564.55,
            "text": " is when context input was removed from the CatBot.",
            "tokens": [
                50392,
                307,
                562,
                4319,
                4846,
                390,
                7261,
                490,
                264,
                9565,
                33,
                310,
                13,
                50648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23330268906612023,
            "compression_ratio": 1.5827814569536425,
            "no_speech_prob": 0.08685302734375,
            "words": [
                {
                    "start": 2558.06,
                    "end": 2558.59,
                    "word": " is",
                    "probability": 0.1368408203125
                },
                {
                    "start": 2558.59,
                    "end": 2559.55,
                    "word": " when",
                    "probability": 0.83740234375
                },
                {
                    "start": 2559.55,
                    "end": 2560.95,
                    "word": " context",
                    "probability": 0.603515625
                },
                {
                    "start": 2560.95,
                    "end": 2561.45,
                    "word": " input",
                    "probability": 0.876953125
                },
                {
                    "start": 2561.45,
                    "end": 2561.99,
                    "word": " was",
                    "probability": 0.88720703125
                },
                {
                    "start": 2561.99,
                    "end": 2562.43,
                    "word": " removed",
                    "probability": 0.75390625
                },
                {
                    "start": 2562.43,
                    "end": 2562.79,
                    "word": " from",
                    "probability": 0.8125
                },
                {
                    "start": 2562.79,
                    "end": 2562.97,
                    "word": " the",
                    "probability": 0.740234375
                },
                {
                    "start": 2562.97,
                    "end": 2563.49,
                    "word": " CatBot",
                    "probability": 0.586273193359375
                },
                {
                    "start": 2563.49,
                    "end": 2564.55,
                    "word": ".",
                    "probability": 0.876953125
                }
            ]
        },
        {
            "id": 361,
            "seek": 250979,
            "start": 2565.01,
            "end": 2573.53,
            "text": " We can see significant performance drop when we remove either adapter or context input,",
            "tokens": [
                50740,
                492,
                393,
                536,
                4776,
                3389,
                3270,
                562,
                321,
                4159,
                2139,
                22860,
                420,
                4319,
                4846,
                11,
                51100
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23330268906612023,
            "compression_ratio": 1.5827814569536425,
            "no_speech_prob": 0.08685302734375,
            "words": [
                {
                    "start": 2565.01,
                    "end": 2565.49,
                    "word": " We",
                    "probability": 0.8828125
                },
                {
                    "start": 2565.49,
                    "end": 2565.73,
                    "word": " can",
                    "probability": 0.89990234375
                },
                {
                    "start": 2565.73,
                    "end": 2566.17,
                    "word": " see",
                    "probability": 0.8212890625
                },
                {
                    "start": 2566.17,
                    "end": 2566.91,
                    "word": " significant",
                    "probability": 0.7353515625
                },
                {
                    "start": 2566.91,
                    "end": 2567.73,
                    "word": " performance",
                    "probability": 0.814453125
                },
                {
                    "start": 2567.73,
                    "end": 2568.17,
                    "word": " drop",
                    "probability": 0.81787109375
                },
                {
                    "start": 2568.17,
                    "end": 2569.15,
                    "word": " when",
                    "probability": 0.8759765625
                },
                {
                    "start": 2569.15,
                    "end": 2569.47,
                    "word": " we",
                    "probability": 0.89013671875
                },
                {
                    "start": 2569.47,
                    "end": 2569.99,
                    "word": " remove",
                    "probability": 0.8193359375
                },
                {
                    "start": 2569.99,
                    "end": 2570.53,
                    "word": " either",
                    "probability": 0.83447265625
                },
                {
                    "start": 2570.53,
                    "end": 2571.01,
                    "word": " adapter",
                    "probability": 0.6396484375
                },
                {
                    "start": 2571.01,
                    "end": 2571.41,
                    "word": " or",
                    "probability": 0.9013671875
                },
                {
                    "start": 2571.41,
                    "end": 2571.89,
                    "word": " context",
                    "probability": 0.77978515625
                },
                {
                    "start": 2571.89,
                    "end": 2572.47,
                    "word": " input",
                    "probability": 0.87158203125
                },
                {
                    "start": 2572.47,
                    "end": 2573.53,
                    "word": ",",
                    "probability": 0.39599609375
                }
            ]
        },
        {
            "id": 362,
            "seek": 250979,
            "start": 2573.53,
            "end": 2583.99,
            "text": " which demonstrates we can improve performance by using additional adapters and context rated layers.",
            "tokens": [
                51152,
                597,
                31034,
                321,
                393,
                3470,
                3389,
                538,
                1228,
                4497,
                23169,
                1559,
                293,
                4319,
                22103,
                7914,
                13,
                51624
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23330268906612023,
            "compression_ratio": 1.5827814569536425,
            "no_speech_prob": 0.08685302734375,
            "words": [
                {
                    "start": 2573.53,
                    "end": 2573.87,
                    "word": " which",
                    "probability": 0.916015625
                },
                {
                    "start": 2573.87,
                    "end": 2574.47,
                    "word": " demonstrates",
                    "probability": 0.344482421875
                },
                {
                    "start": 2574.47,
                    "end": 2575.15,
                    "word": " we",
                    "probability": 0.87646484375
                },
                {
                    "start": 2575.15,
                    "end": 2575.35,
                    "word": " can",
                    "probability": 0.8935546875
                },
                {
                    "start": 2575.35,
                    "end": 2575.71,
                    "word": " improve",
                    "probability": 0.71826171875
                },
                {
                    "start": 2575.71,
                    "end": 2576.25,
                    "word": " performance",
                    "probability": 0.58935546875
                },
                {
                    "start": 2576.25,
                    "end": 2576.63,
                    "word": " by",
                    "probability": 0.8984375
                },
                {
                    "start": 2576.63,
                    "end": 2578.15,
                    "word": " using",
                    "probability": 0.80322265625
                },
                {
                    "start": 2578.15,
                    "end": 2579.37,
                    "word": " additional",
                    "probability": 0.8154296875
                },
                {
                    "start": 2579.37,
                    "end": 2580.07,
                    "word": " adapters",
                    "probability": 0.875
                },
                {
                    "start": 2580.07,
                    "end": 2580.37,
                    "word": " and",
                    "probability": 0.884765625
                },
                {
                    "start": 2580.37,
                    "end": 2581.01,
                    "word": " context",
                    "probability": 0.73779296875
                },
                {
                    "start": 2581.01,
                    "end": 2581.45,
                    "word": " rated",
                    "probability": 0.0552978515625
                },
                {
                    "start": 2581.45,
                    "end": 2582.93,
                    "word": " layers",
                    "probability": 0.86083984375
                },
                {
                    "start": 2582.93,
                    "end": 2583.99,
                    "word": ".",
                    "probability": 0.98046875
                }
            ]
        },
        {
            "id": 363,
            "seek": 253585,
            "start": 2587.76,
            "end": 2598.01,
            "text": " Next, we compare the performance of three models with targeted BC samples.",
            "tokens": [
                50412,
                3087,
                11,
                321,
                6794,
                264,
                3389,
                295,
                1045,
                5245,
                365,
                15045,
                14359,
                10938,
                13,
                50860
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24894831845393547,
            "compression_ratio": 1.5902777777777777,
            "no_speech_prob": 0.0021190643310546875,
            "words": [
                {
                    "start": 2587.76,
                    "end": 2588.33,
                    "word": " Next",
                    "probability": 0.71533203125
                },
                {
                    "start": 2588.33,
                    "end": 2588.47,
                    "word": ",",
                    "probability": 0.8037109375
                },
                {
                    "start": 2588.47,
                    "end": 2588.63,
                    "word": " we",
                    "probability": 0.88525390625
                },
                {
                    "start": 2588.63,
                    "end": 2589.25,
                    "word": " compare",
                    "probability": 0.71044921875
                },
                {
                    "start": 2589.25,
                    "end": 2590.03,
                    "word": " the",
                    "probability": 0.7470703125
                },
                {
                    "start": 2590.03,
                    "end": 2590.73,
                    "word": " performance",
                    "probability": 0.82958984375
                },
                {
                    "start": 2590.73,
                    "end": 2591.47,
                    "word": " of",
                    "probability": 0.8798828125
                },
                {
                    "start": 2591.47,
                    "end": 2591.99,
                    "word": " three",
                    "probability": 0.60009765625
                },
                {
                    "start": 2591.99,
                    "end": 2592.47,
                    "word": " models",
                    "probability": 0.77490234375
                },
                {
                    "start": 2592.47,
                    "end": 2594.03,
                    "word": " with",
                    "probability": 0.775390625
                },
                {
                    "start": 2594.03,
                    "end": 2595.05,
                    "word": " targeted",
                    "probability": 0.76708984375
                },
                {
                    "start": 2595.05,
                    "end": 2595.53,
                    "word": " BC",
                    "probability": 0.6416015625
                },
                {
                    "start": 2595.53,
                    "end": 2596.95,
                    "word": " samples",
                    "probability": 0.6044921875
                },
                {
                    "start": 2596.95,
                    "end": 2598.01,
                    "word": ".",
                    "probability": 0.94482421875
                }
            ]
        },
        {
            "id": 364,
            "seek": 253585,
            "start": 2598.01,
            "end": 2605.45,
            "text": " We assigned high sample rate for BC samples, and we achieved high performance for detecting",
            "tokens": [
                50924,
                492,
                13279,
                1090,
                6889,
                3314,
                337,
                14359,
                10938,
                11,
                293,
                321,
                11042,
                1090,
                3389,
                337,
                40237,
                51308
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24894831845393547,
            "compression_ratio": 1.5902777777777777,
            "no_speech_prob": 0.0021190643310546875,
            "words": [
                {
                    "start": 2598.01,
                    "end": 2598.47,
                    "word": " We",
                    "probability": 0.88427734375
                },
                {
                    "start": 2598.47,
                    "end": 2598.97,
                    "word": " assigned",
                    "probability": 0.62744140625
                },
                {
                    "start": 2598.97,
                    "end": 2599.31,
                    "word": " high",
                    "probability": 0.8193359375
                },
                {
                    "start": 2599.31,
                    "end": 2599.67,
                    "word": " sample",
                    "probability": 0.76318359375
                },
                {
                    "start": 2599.67,
                    "end": 2599.95,
                    "word": " rate",
                    "probability": 0.32177734375
                },
                {
                    "start": 2599.95,
                    "end": 2600.19,
                    "word": " for",
                    "probability": 0.89306640625
                },
                {
                    "start": 2600.19,
                    "end": 2600.55,
                    "word": " BC",
                    "probability": 0.89111328125
                },
                {
                    "start": 2600.55,
                    "end": 2601.19,
                    "word": " samples",
                    "probability": 0.74658203125
                },
                {
                    "start": 2601.19,
                    "end": 2601.51,
                    "word": ",",
                    "probability": 0.334228515625
                },
                {
                    "start": 2601.51,
                    "end": 2601.65,
                    "word": " and",
                    "probability": 0.87158203125
                },
                {
                    "start": 2601.65,
                    "end": 2602.01,
                    "word": " we",
                    "probability": 0.8466796875
                },
                {
                    "start": 2602.01,
                    "end": 2603.13,
                    "word": " achieved",
                    "probability": 0.72705078125
                },
                {
                    "start": 2603.13,
                    "end": 2603.71,
                    "word": " high",
                    "probability": 0.87353515625
                },
                {
                    "start": 2603.71,
                    "end": 2604.29,
                    "word": " performance",
                    "probability": 0.7900390625
                },
                {
                    "start": 2604.29,
                    "end": 2604.69,
                    "word": " for",
                    "probability": 0.884765625
                },
                {
                    "start": 2604.69,
                    "end": 2605.45,
                    "word": " detecting",
                    "probability": 0.7958984375
                }
            ]
        },
        {
            "id": 365,
            "seek": 253585,
            "start": 2605.45,
            "end": 2613.05,
            "text": " those BC samples and the CatBot above the two baseline models.",
            "tokens": [
                51308,
                729,
                14359,
                10938,
                293,
                264,
                9565,
                33,
                310,
                3673,
                264,
                732,
                20518,
                5245,
                13,
                51604
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24894831845393547,
            "compression_ratio": 1.5902777777777777,
            "no_speech_prob": 0.0021190643310546875,
            "words": [
                {
                    "start": 2605.45,
                    "end": 2606.29,
                    "word": " those",
                    "probability": 0.6826171875
                },
                {
                    "start": 2606.29,
                    "end": 2606.65,
                    "word": " BC",
                    "probability": 0.97216796875
                },
                {
                    "start": 2606.65,
                    "end": 2607.21,
                    "word": " samples",
                    "probability": 0.7607421875
                },
                {
                    "start": 2607.21,
                    "end": 2607.91,
                    "word": " and",
                    "probability": 0.499755859375
                },
                {
                    "start": 2607.91,
                    "end": 2608.83,
                    "word": " the",
                    "probability": 0.71728515625
                },
                {
                    "start": 2608.83,
                    "end": 2609.41,
                    "word": " CatBot",
                    "probability": 0.4275716145833333
                },
                {
                    "start": 2609.41,
                    "end": 2609.83,
                    "word": " above",
                    "probability": 0.208740234375
                },
                {
                    "start": 2609.83,
                    "end": 2610.33,
                    "word": " the",
                    "probability": 0.75048828125
                },
                {
                    "start": 2610.33,
                    "end": 2610.59,
                    "word": " two",
                    "probability": 0.802734375
                },
                {
                    "start": 2610.59,
                    "end": 2611.21,
                    "word": " baseline",
                    "probability": 0.9111328125
                },
                {
                    "start": 2611.21,
                    "end": 2611.91,
                    "word": " models",
                    "probability": 0.8046875
                },
                {
                    "start": 2611.91,
                    "end": 2613.05,
                    "word": ".",
                    "probability": 0.98193359375
                }
            ]
        },
        {
            "id": 366,
            "seek": 256487,
            "start": 2615.46,
            "end": 2624.42,
            "text": " Next, we compare performance for phishing emails. We divide the phishing emails into two groups,",
            "tokens": [
                50404,
                3087,
                11,
                321,
                6794,
                3389,
                337,
                903,
                3807,
                12524,
                13,
                492,
                9845,
                264,
                903,
                3807,
                12524,
                666,
                732,
                3935,
                11,
                50772
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23154439010481903,
            "compression_ratio": 1.7080745341614907,
            "no_speech_prob": 0.01433563232421875,
            "words": [
                {
                    "start": 2615.46,
                    "end": 2615.86,
                    "word": " Next",
                    "probability": 0.6103515625
                },
                {
                    "start": 2615.86,
                    "end": 2616.04,
                    "word": ",",
                    "probability": 0.703125
                },
                {
                    "start": 2616.04,
                    "end": 2616.14,
                    "word": " we",
                    "probability": 0.86328125
                },
                {
                    "start": 2616.14,
                    "end": 2616.72,
                    "word": " compare",
                    "probability": 0.7333984375
                },
                {
                    "start": 2616.72,
                    "end": 2617.4,
                    "word": " performance",
                    "probability": 0.63037109375
                },
                {
                    "start": 2617.4,
                    "end": 2617.82,
                    "word": " for",
                    "probability": 0.79052734375
                },
                {
                    "start": 2617.82,
                    "end": 2618.6,
                    "word": " phishing",
                    "probability": 0.87939453125
                },
                {
                    "start": 2618.6,
                    "end": 2618.98,
                    "word": " emails",
                    "probability": 0.7626953125
                },
                {
                    "start": 2618.98,
                    "end": 2619.78,
                    "word": ".",
                    "probability": 0.91796875
                },
                {
                    "start": 2620.12,
                    "end": 2620.34,
                    "word": " We",
                    "probability": 0.87060546875
                },
                {
                    "start": 2620.34,
                    "end": 2620.78,
                    "word": " divide",
                    "probability": 0.7158203125
                },
                {
                    "start": 2620.78,
                    "end": 2620.98,
                    "word": " the",
                    "probability": 0.443359375
                },
                {
                    "start": 2620.98,
                    "end": 2621.3,
                    "word": " phishing",
                    "probability": 0.89599609375
                },
                {
                    "start": 2621.3,
                    "end": 2621.68,
                    "word": " emails",
                    "probability": 0.75048828125
                },
                {
                    "start": 2621.68,
                    "end": 2622.06,
                    "word": " into",
                    "probability": 0.7294921875
                },
                {
                    "start": 2622.06,
                    "end": 2622.4,
                    "word": " two",
                    "probability": 0.5908203125
                },
                {
                    "start": 2622.4,
                    "end": 2622.98,
                    "word": " groups",
                    "probability": 0.787109375
                },
                {
                    "start": 2622.98,
                    "end": 2624.42,
                    "word": ",",
                    "probability": 0.475341796875
                }
            ]
        },
        {
            "id": 367,
            "seek": 256487,
            "start": 2624.42,
            "end": 2634.76,
            "text": " English and non-English emails. Our CatBot outperformed for English and non-English emails.",
            "tokens": [
                50844,
                3669,
                293,
                2107,
                12,
                31254,
                1933,
                12524,
                13,
                2621,
                9565,
                33,
                310,
                484,
                610,
                22892,
                337,
                3669,
                293,
                2107,
                12,
                31254,
                1933,
                12524,
                13,
                51348
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23154439010481903,
            "compression_ratio": 1.7080745341614907,
            "no_speech_prob": 0.01433563232421875,
            "words": [
                {
                    "start": 2624.42,
                    "end": 2624.74,
                    "word": " English",
                    "probability": 0.87158203125
                },
                {
                    "start": 2624.74,
                    "end": 2625.06,
                    "word": " and",
                    "probability": 0.873046875
                },
                {
                    "start": 2625.06,
                    "end": 2625.38,
                    "word": " non",
                    "probability": 0.87451171875
                },
                {
                    "start": 2625.38,
                    "end": 2625.86,
                    "word": "-English",
                    "probability": 0.9453125
                },
                {
                    "start": 2625.86,
                    "end": 2626.52,
                    "word": " emails",
                    "probability": 0.6826171875
                },
                {
                    "start": 2626.52,
                    "end": 2627.32,
                    "word": ".",
                    "probability": 0.96337890625
                },
                {
                    "start": 2628.14,
                    "end": 2628.7,
                    "word": " Our",
                    "probability": 0.85498046875
                },
                {
                    "start": 2628.7,
                    "end": 2629.94,
                    "word": " CatBot",
                    "probability": 0.3739166259765625
                },
                {
                    "start": 2629.94,
                    "end": 2630.9,
                    "word": " outperformed",
                    "probability": 0.5593465169270834
                },
                {
                    "start": 2630.9,
                    "end": 2632.48,
                    "word": " for",
                    "probability": 0.8310546875
                },
                {
                    "start": 2632.48,
                    "end": 2632.88,
                    "word": " English",
                    "probability": 0.8857421875
                },
                {
                    "start": 2632.88,
                    "end": 2633.2,
                    "word": " and",
                    "probability": 0.88232421875
                },
                {
                    "start": 2633.2,
                    "end": 2633.48,
                    "word": " non",
                    "probability": 0.919921875
                },
                {
                    "start": 2633.48,
                    "end": 2633.84,
                    "word": "-English",
                    "probability": 0.95849609375
                },
                {
                    "start": 2633.84,
                    "end": 2634.36,
                    "word": " emails",
                    "probability": 0.7216796875
                },
                {
                    "start": 2634.36,
                    "end": 2634.76,
                    "word": ".",
                    "probability": 0.64404296875
                }
            ]
        },
        {
            "id": 368,
            "seek": 256487,
            "start": 2636.0,
            "end": 2643.72,
            "text": " And our model was based on the multilingual bot, so we can see significant performance",
            "tokens": [
                51424,
                400,
                527,
                2316,
                390,
                2361,
                322,
                264,
                2120,
                38219,
                10592,
                11,
                370,
                321,
                393,
                536,
                4776,
                3389,
                51832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23154439010481903,
            "compression_ratio": 1.7080745341614907,
            "no_speech_prob": 0.01433563232421875,
            "words": [
                {
                    "start": 2636.0,
                    "end": 2636.28,
                    "word": " And",
                    "probability": 0.410888671875
                },
                {
                    "start": 2636.28,
                    "end": 2637.38,
                    "word": " our",
                    "probability": 0.78076171875
                },
                {
                    "start": 2637.38,
                    "end": 2637.72,
                    "word": " model",
                    "probability": 0.7880859375
                },
                {
                    "start": 2637.72,
                    "end": 2638.12,
                    "word": " was",
                    "probability": 0.85888671875
                },
                {
                    "start": 2638.12,
                    "end": 2638.68,
                    "word": " based",
                    "probability": 0.81787109375
                },
                {
                    "start": 2638.68,
                    "end": 2638.96,
                    "word": " on",
                    "probability": 0.90087890625
                },
                {
                    "start": 2638.96,
                    "end": 2639.14,
                    "word": " the",
                    "probability": 0.7353515625
                },
                {
                    "start": 2639.14,
                    "end": 2639.74,
                    "word": " multilingual",
                    "probability": 0.6292724609375
                },
                {
                    "start": 2639.74,
                    "end": 2640.26,
                    "word": " bot",
                    "probability": 0.70361328125
                },
                {
                    "start": 2640.26,
                    "end": 2640.56,
                    "word": ",",
                    "probability": 0.498779296875
                },
                {
                    "start": 2640.56,
                    "end": 2640.8,
                    "word": " so",
                    "probability": 0.89453125
                },
                {
                    "start": 2640.8,
                    "end": 2641.1,
                    "word": " we",
                    "probability": 0.90625
                },
                {
                    "start": 2641.1,
                    "end": 2641.3,
                    "word": " can",
                    "probability": 0.89599609375
                },
                {
                    "start": 2641.3,
                    "end": 2641.68,
                    "word": " see",
                    "probability": 0.83203125
                },
                {
                    "start": 2641.68,
                    "end": 2642.74,
                    "word": " significant",
                    "probability": 0.740234375
                },
                {
                    "start": 2642.74,
                    "end": 2643.72,
                    "word": " performance",
                    "probability": 0.81689453125
                }
            ]
        },
        {
            "id": 369,
            "seek": 259299,
            "start": 2643.72,
            "end": 2654.97,
            "text": " when we use the model for detecting non-English emails, when it compares with simple LST model.",
            "tokens": [
                50380,
                562,
                321,
                764,
                264,
                2316,
                337,
                40237,
                2107,
                12,
                31254,
                1933,
                12524,
                11,
                562,
                309,
                38334,
                365,
                2199,
                441,
                6840,
                2316,
                13,
                50816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3215579692868219,
            "compression_ratio": 1.580110497237569,
            "no_speech_prob": 0.0021648406982421875,
            "words": [
                {
                    "start": 2643.72,
                    "end": 2644.38,
                    "word": " when",
                    "probability": 0.217041015625
                },
                {
                    "start": 2644.38,
                    "end": 2644.62,
                    "word": " we",
                    "probability": 0.69921875
                },
                {
                    "start": 2644.62,
                    "end": 2645.66,
                    "word": " use",
                    "probability": 0.65283203125
                },
                {
                    "start": 2645.66,
                    "end": 2645.84,
                    "word": " the",
                    "probability": 0.72705078125
                },
                {
                    "start": 2645.84,
                    "end": 2646.12,
                    "word": " model",
                    "probability": 0.7392578125
                },
                {
                    "start": 2646.12,
                    "end": 2646.54,
                    "word": " for",
                    "probability": 0.5458984375
                },
                {
                    "start": 2646.54,
                    "end": 2646.92,
                    "word": " detecting",
                    "probability": 0.7255859375
                },
                {
                    "start": 2646.92,
                    "end": 2647.26,
                    "word": " non",
                    "probability": 0.85986328125
                },
                {
                    "start": 2647.26,
                    "end": 2647.62,
                    "word": "-English",
                    "probability": 0.8369140625
                },
                {
                    "start": 2647.62,
                    "end": 2648.04,
                    "word": " emails",
                    "probability": 0.439453125
                },
                {
                    "start": 2648.04,
                    "end": 2649.16,
                    "word": ",",
                    "probability": 0.30810546875
                },
                {
                    "start": 2650.53,
                    "end": 2651.51,
                    "word": " when",
                    "probability": 0.79296875
                },
                {
                    "start": 2651.51,
                    "end": 2651.75,
                    "word": " it",
                    "probability": 0.84814453125
                },
                {
                    "start": 2651.75,
                    "end": 2652.09,
                    "word": " compares",
                    "probability": 0.73779296875
                },
                {
                    "start": 2652.09,
                    "end": 2652.43,
                    "word": " with",
                    "probability": 0.80419921875
                },
                {
                    "start": 2652.43,
                    "end": 2653.27,
                    "word": " simple",
                    "probability": 0.59521484375
                },
                {
                    "start": 2653.27,
                    "end": 2653.73,
                    "word": " LST",
                    "probability": 0.925537109375
                },
                {
                    "start": 2653.73,
                    "end": 2654.15,
                    "word": " model",
                    "probability": 0.73828125
                },
                {
                    "start": 2654.15,
                    "end": 2654.97,
                    "word": ".",
                    "probability": 0.935546875
                }
            ]
        },
        {
            "id": 370,
            "seek": 259299,
            "start": 2656.57,
            "end": 2664.33,
            "text": " Next, we compare the inference speed. This bot has six transformer blocks, and the CatBot has",
            "tokens": [
                50960,
                3087,
                11,
                321,
                6794,
                264,
                38253,
                3073,
                13,
                639,
                10592,
                575,
                2309,
                31782,
                8474,
                11,
                293,
                264,
                9565,
                33,
                310,
                575,
                51320
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3215579692868219,
            "compression_ratio": 1.580110497237569,
            "no_speech_prob": 0.0021648406982421875,
            "words": [
                {
                    "start": 2656.57,
                    "end": 2657.39,
                    "word": " Next",
                    "probability": 0.87939453125
                },
                {
                    "start": 2657.39,
                    "end": 2657.49,
                    "word": ",",
                    "probability": 0.7451171875
                },
                {
                    "start": 2657.49,
                    "end": 2657.61,
                    "word": " we",
                    "probability": 0.89892578125
                },
                {
                    "start": 2657.61,
                    "end": 2658.13,
                    "word": " compare",
                    "probability": 0.66845703125
                },
                {
                    "start": 2658.13,
                    "end": 2658.57,
                    "word": " the",
                    "probability": 0.7626953125
                },
                {
                    "start": 2658.57,
                    "end": 2658.97,
                    "word": " inference",
                    "probability": 0.826171875
                },
                {
                    "start": 2658.97,
                    "end": 2659.51,
                    "word": " speed",
                    "probability": 0.8447265625
                },
                {
                    "start": 2659.51,
                    "end": 2660.33,
                    "word": ".",
                    "probability": 0.9580078125
                },
                {
                    "start": 2660.55,
                    "end": 2660.69,
                    "word": " This",
                    "probability": 0.1441650390625
                },
                {
                    "start": 2660.69,
                    "end": 2660.99,
                    "word": " bot",
                    "probability": 0.005680084228515625
                },
                {
                    "start": 2660.99,
                    "end": 2661.35,
                    "word": " has",
                    "probability": 0.888671875
                },
                {
                    "start": 2661.35,
                    "end": 2661.65,
                    "word": " six",
                    "probability": 0.4990234375
                },
                {
                    "start": 2661.65,
                    "end": 2662.23,
                    "word": " transformer",
                    "probability": 0.67236328125
                },
                {
                    "start": 2662.23,
                    "end": 2662.71,
                    "word": " blocks",
                    "probability": 0.86083984375
                },
                {
                    "start": 2662.71,
                    "end": 2662.91,
                    "word": ",",
                    "probability": 0.345947265625
                },
                {
                    "start": 2662.91,
                    "end": 2663.15,
                    "word": " and",
                    "probability": 0.896484375
                },
                {
                    "start": 2663.15,
                    "end": 2663.55,
                    "word": " the",
                    "probability": 0.6201171875
                },
                {
                    "start": 2663.55,
                    "end": 2663.99,
                    "word": " CatBot",
                    "probability": 0.45556640625
                },
                {
                    "start": 2663.99,
                    "end": 2664.33,
                    "word": " has",
                    "probability": 0.90185546875
                }
            ]
        },
        {
            "id": 371,
            "seek": 259299,
            "start": 2664.33,
            "end": 2673.21,
            "text": " three transformer blocks, so we achieve two times speed up in inference time when we measure the",
            "tokens": [
                51320,
                1045,
                31782,
                8474,
                11,
                370,
                321,
                4584,
                732,
                1413,
                3073,
                493,
                294,
                38253,
                565,
                562,
                321,
                3481,
                264,
                51764
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3215579692868219,
            "compression_ratio": 1.580110497237569,
            "no_speech_prob": 0.0021648406982421875,
            "words": [
                {
                    "start": 2664.33,
                    "end": 2664.67,
                    "word": " three",
                    "probability": 0.84375
                },
                {
                    "start": 2664.67,
                    "end": 2665.21,
                    "word": " transformer",
                    "probability": 0.7958984375
                },
                {
                    "start": 2665.21,
                    "end": 2665.65,
                    "word": " blocks",
                    "probability": 0.8330078125
                },
                {
                    "start": 2665.65,
                    "end": 2666.29,
                    "word": ",",
                    "probability": 0.1929931640625
                },
                {
                    "start": 2666.29,
                    "end": 2666.63,
                    "word": " so",
                    "probability": 0.876953125
                },
                {
                    "start": 2666.63,
                    "end": 2667.21,
                    "word": " we",
                    "probability": 0.876953125
                },
                {
                    "start": 2667.21,
                    "end": 2667.69,
                    "word": " achieve",
                    "probability": 0.65576171875
                },
                {
                    "start": 2667.69,
                    "end": 2668.21,
                    "word": " two",
                    "probability": 0.54638671875
                },
                {
                    "start": 2668.21,
                    "end": 2668.71,
                    "word": " times",
                    "probability": 0.77490234375
                },
                {
                    "start": 2668.71,
                    "end": 2669.63,
                    "word": " speed",
                    "probability": 0.70068359375
                },
                {
                    "start": 2669.63,
                    "end": 2669.85,
                    "word": " up",
                    "probability": 0.27587890625
                },
                {
                    "start": 2669.85,
                    "end": 2670.47,
                    "word": " in",
                    "probability": 0.86572265625
                },
                {
                    "start": 2670.47,
                    "end": 2671.75,
                    "word": " inference",
                    "probability": 0.82177734375
                },
                {
                    "start": 2671.75,
                    "end": 2672.09,
                    "word": " time",
                    "probability": 0.802734375
                },
                {
                    "start": 2672.09,
                    "end": 2672.39,
                    "word": " when",
                    "probability": 0.75439453125
                },
                {
                    "start": 2672.39,
                    "end": 2672.67,
                    "word": " we",
                    "probability": 0.90234375
                },
                {
                    "start": 2672.67,
                    "end": 2672.93,
                    "word": " measure",
                    "probability": 0.321044921875
                },
                {
                    "start": 2672.93,
                    "end": 2673.21,
                    "word": " the",
                    "probability": 0.7607421875
                }
            ]
        },
        {
            "id": 372,
            "seek": 261445,
            "start": 2673.21,
            "end": 2686.09,
            "text": " performance on a CPU machine. As the number of blocks decrease, the inference time can be reduced.",
            "tokens": [
                50364,
                3389,
                322,
                257,
                13199,
                3479,
                13,
                1018,
                264,
                1230,
                295,
                8474,
                11514,
                11,
                264,
                38253,
                565,
                393,
                312,
                9212,
                13,
                50988
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24182412236235862,
            "compression_ratio": 1.2868217054263567,
            "no_speech_prob": 0.004085540771484375,
            "words": [
                {
                    "start": 2673.21,
                    "end": 2673.63,
                    "word": " performance",
                    "probability": 0.2100830078125
                },
                {
                    "start": 2673.63,
                    "end": 2674.05,
                    "word": " on",
                    "probability": 0.810546875
                },
                {
                    "start": 2674.47,
                    "end": 2674.95,
                    "word": " a CPU",
                    "probability": 0.80029296875
                },
                {
                    "start": 2674.95,
                    "end": 2675.43,
                    "word": " machine",
                    "probability": 0.6748046875
                },
                {
                    "start": 2675.43,
                    "end": 2676.27,
                    "word": ".",
                    "probability": 0.93701171875
                },
                {
                    "start": 2676.73,
                    "end": 2677.57,
                    "word": " As",
                    "probability": 0.87939453125
                },
                {
                    "start": 2677.57,
                    "end": 2677.71,
                    "word": " the",
                    "probability": 0.68798828125
                },
                {
                    "start": 2677.71,
                    "end": 2677.89,
                    "word": " number",
                    "probability": 0.87646484375
                },
                {
                    "start": 2677.89,
                    "end": 2678.13,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 2678.13,
                    "end": 2678.47,
                    "word": " blocks",
                    "probability": 0.7734375
                },
                {
                    "start": 2678.47,
                    "end": 2679.45,
                    "word": " decrease",
                    "probability": 0.4287109375
                },
                {
                    "start": 2679.45,
                    "end": 2681.05,
                    "word": ",",
                    "probability": 0.923828125
                },
                {
                    "start": 2681.05,
                    "end": 2681.37,
                    "word": " the",
                    "probability": 0.77392578125
                },
                {
                    "start": 2681.37,
                    "end": 2682.03,
                    "word": " inference",
                    "probability": 0.75634765625
                },
                {
                    "start": 2682.03,
                    "end": 2683.51,
                    "word": " time",
                    "probability": 0.7734375
                },
                {
                    "start": 2683.51,
                    "end": 2685.03,
                    "word": " can",
                    "probability": 0.88916015625
                },
                {
                    "start": 2685.03,
                    "end": 2685.21,
                    "word": " be",
                    "probability": 0.9091796875
                },
                {
                    "start": 2685.21,
                    "end": 2685.67,
                    "word": " reduced",
                    "probability": 0.783203125
                },
                {
                    "start": 2685.67,
                    "end": 2686.09,
                    "word": ".",
                    "probability": 0.978515625
                }
            ]
        },
        {
            "id": 373,
            "seek": 261445,
            "start": 2689.56,
            "end": 2691.9,
            "text": " Next, model size.",
            "tokens": [
                51080,
                3087,
                11,
                2316,
                2744,
                13,
                51140
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24182412236235862,
            "compression_ratio": 1.2868217054263567,
            "no_speech_prob": 0.004085540771484375,
            "words": [
                {
                    "start": 2689.56,
                    "end": 2690.22,
                    "word": " Next",
                    "probability": 0.740234375
                },
                {
                    "start": 2690.22,
                    "end": 2690.36,
                    "word": ",",
                    "probability": 0.79931640625
                },
                {
                    "start": 2690.36,
                    "end": 2690.58,
                    "word": " model",
                    "probability": 0.67724609375
                },
                {
                    "start": 2690.58,
                    "end": 2691.06,
                    "word": " size",
                    "probability": 0.794921875
                },
                {
                    "start": 2691.06,
                    "end": 2691.9,
                    "word": ".",
                    "probability": 0.9404296875
                }
            ]
        },
        {
            "id": 374,
            "seek": 261445,
            "start": 2693.82,
            "end": 2697.0,
            "text": " For comparison, we divide the model size into two",
            "tokens": [
                51280,
                1171,
                9660,
                11,
                321,
                9845,
                264,
                2316,
                2744,
                666,
                732,
                51436
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24182412236235862,
            "compression_ratio": 1.2868217054263567,
            "no_speech_prob": 0.004085540771484375,
            "words": [
                {
                    "start": 2693.82,
                    "end": 2694.14,
                    "word": " For",
                    "probability": 0.90771484375
                },
                {
                    "start": 2694.14,
                    "end": 2694.58,
                    "word": " comparison",
                    "probability": 0.74072265625
                },
                {
                    "start": 2694.58,
                    "end": 2694.76,
                    "word": ",",
                    "probability": 0.9423828125
                },
                {
                    "start": 2694.76,
                    "end": 2694.98,
                    "word": " we",
                    "probability": 0.89794921875
                },
                {
                    "start": 2694.98,
                    "end": 2695.44,
                    "word": " divide",
                    "probability": 0.7578125
                },
                {
                    "start": 2695.44,
                    "end": 2695.6,
                    "word": " the",
                    "probability": 0.6552734375
                },
                {
                    "start": 2695.6,
                    "end": 2695.82,
                    "word": " model",
                    "probability": 0.78564453125
                },
                {
                    "start": 2695.82,
                    "end": 2696.2,
                    "word": " size",
                    "probability": 0.8154296875
                },
                {
                    "start": 2696.2,
                    "end": 2696.58,
                    "word": " into",
                    "probability": 0.72900390625
                },
                {
                    "start": 2696.58,
                    "end": 2697.0,
                    "word": " two",
                    "probability": 0.77783203125
                }
            ]
        },
        {
            "id": 375,
            "seek": 264033,
            "start": 2698.4,
            "end": 2703.18,
            "text": " parts, embedding and transformer blocks.",
            "tokens": [
                50396,
                971,
                1373,
                11,
                12240,
                3584,
                293,
                31782,
                8474,
                13,
                50560
            ],
            "temperature": 0.0,
            "avg_logprob": -0.39787946535008295,
            "compression_ratio": 1.8583333333333334,
            "no_speech_prob": 0.0211181640625,
            "words": [
                {
                    "start": 2698.4,
                    "end": 2699.44,
                    "word": " parts",
                    "probability": 0.06731033325195312
                },
                {
                    "start": 2699.44,
                    "end": 2699.8,
                    "word": ",",
                    "probability": 0.15478515625
                },
                {
                    "start": 2699.8,
                    "end": 2700.48,
                    "word": " embedding",
                    "probability": 0.8154296875
                },
                {
                    "start": 2700.48,
                    "end": 2700.78,
                    "word": " and",
                    "probability": 0.83544921875
                },
                {
                    "start": 2700.78,
                    "end": 2701.46,
                    "word": " transformer",
                    "probability": 0.77783203125
                },
                {
                    "start": 2701.46,
                    "end": 2702.14,
                    "word": " blocks",
                    "probability": 0.86962890625
                },
                {
                    "start": 2702.14,
                    "end": 2703.18,
                    "word": ".",
                    "probability": 0.95458984375
                }
            ]
        },
        {
            "id": 376,
            "seek": 264033,
            "start": 2706.19,
            "end": 2716.99,
            "text": " The DistilBot has six transformers and has 92 million parameters for embedding and 42 million",
            "tokens": [
                50656,
                440,
                9840,
                388,
                33,
                310,
                575,
                2309,
                4088,
                433,
                293,
                575,
                28225,
                2459,
                9834,
                337,
                12240,
                3584,
                293,
                14034,
                2459,
                51204
            ],
            "temperature": 0.0,
            "avg_logprob": -0.39787946535008295,
            "compression_ratio": 1.8583333333333334,
            "no_speech_prob": 0.0211181640625,
            "words": [
                {
                    "start": 2706.19,
                    "end": 2706.33,
                    "word": " The",
                    "probability": 0.294921875
                },
                {
                    "start": 2706.33,
                    "end": 2706.89,
                    "word": " DistilBot",
                    "probability": 0.44219970703125
                },
                {
                    "start": 2706.89,
                    "end": 2707.21,
                    "word": " has",
                    "probability": 0.8857421875
                },
                {
                    "start": 2707.21,
                    "end": 2707.61,
                    "word": " six",
                    "probability": 0.455078125
                },
                {
                    "start": 2707.61,
                    "end": 2708.59,
                    "word": " transformers",
                    "probability": 0.88916015625
                },
                {
                    "start": 2708.59,
                    "end": 2710.41,
                    "word": " and",
                    "probability": 0.7724609375
                },
                {
                    "start": 2710.41,
                    "end": 2711.45,
                    "word": " has",
                    "probability": 0.8603515625
                },
                {
                    "start": 2711.45,
                    "end": 2713.49,
                    "word": " 92",
                    "probability": 0.78369140625
                },
                {
                    "start": 2713.49,
                    "end": 2713.95,
                    "word": " million",
                    "probability": 0.74853515625
                },
                {
                    "start": 2713.95,
                    "end": 2714.53,
                    "word": " parameters",
                    "probability": 0.8720703125
                },
                {
                    "start": 2714.53,
                    "end": 2714.99,
                    "word": " for",
                    "probability": 0.89599609375
                },
                {
                    "start": 2714.99,
                    "end": 2715.53,
                    "word": " embedding",
                    "probability": 0.91357421875
                },
                {
                    "start": 2715.53,
                    "end": 2715.85,
                    "word": " and",
                    "probability": 0.87060546875
                },
                {
                    "start": 2715.85,
                    "end": 2716.55,
                    "word": " 42",
                    "probability": 0.83935546875
                },
                {
                    "start": 2716.55,
                    "end": 2716.99,
                    "word": " million",
                    "probability": 0.77392578125
                }
            ]
        },
        {
            "id": 377,
            "seek": 264033,
            "start": 2716.99,
            "end": 2725.01,
            "text": " parameters for transformer blocks. DistilBot and the CatBot have six transformer blocks.",
            "tokens": [
                51204,
                9834,
                337,
                31782,
                8474,
                13,
                9840,
                388,
                33,
                310,
                293,
                264,
                9565,
                33,
                310,
                362,
                2309,
                31782,
                8474,
                13,
                51596
            ],
            "temperature": 0.0,
            "avg_logprob": -0.39787946535008295,
            "compression_ratio": 1.8583333333333334,
            "no_speech_prob": 0.0211181640625,
            "words": [
                {
                    "start": 2716.99,
                    "end": 2717.59,
                    "word": " parameters",
                    "probability": 0.89599609375
                },
                {
                    "start": 2717.59,
                    "end": 2718.01,
                    "word": " for",
                    "probability": 0.8984375
                },
                {
                    "start": 2718.01,
                    "end": 2718.57,
                    "word": " transformer",
                    "probability": 0.8408203125
                },
                {
                    "start": 2718.57,
                    "end": 2719.05,
                    "word": " blocks",
                    "probability": 0.84619140625
                },
                {
                    "start": 2719.05,
                    "end": 2720.09,
                    "word": ".",
                    "probability": 0.9287109375
                },
                {
                    "start": 2720.73,
                    "end": 2721.75,
                    "word": " DistilBot",
                    "probability": 0.7218017578125
                },
                {
                    "start": 2721.75,
                    "end": 2722.13,
                    "word": " and",
                    "probability": 0.78515625
                },
                {
                    "start": 2722.13,
                    "end": 2722.49,
                    "word": " the",
                    "probability": 0.491943359375
                },
                {
                    "start": 2722.49,
                    "end": 2722.91,
                    "word": " CatBot",
                    "probability": 0.79345703125
                },
                {
                    "start": 2722.91,
                    "end": 2723.17,
                    "word": " have",
                    "probability": 0.323486328125
                },
                {
                    "start": 2723.17,
                    "end": 2723.51,
                    "word": " six",
                    "probability": 0.8525390625
                },
                {
                    "start": 2723.51,
                    "end": 2724.07,
                    "word": " transformer",
                    "probability": 0.765625
                },
                {
                    "start": 2724.07,
                    "end": 2724.57,
                    "word": " blocks",
                    "probability": 0.8369140625
                },
                {
                    "start": 2724.57,
                    "end": 2725.01,
                    "word": ".",
                    "probability": 0.7412109375
                }
            ]
        },
        {
            "id": 378,
            "seek": 267033,
            "start": 2725.01,
            "end": 2731.55,
            "text": " We reuse the same embedding parameters, but we reduce the number of parameters for",
            "tokens": [
                50364,
                492,
                26225,
                264,
                912,
                12240,
                3584,
                9834,
                11,
                457,
                321,
                5407,
                264,
                1230,
                295,
                9834,
                337,
                50696
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2219669137369184,
            "compression_ratio": 1.5494505494505495,
            "no_speech_prob": 0.01506805419921875,
            "words": [
                {
                    "start": 2725.01,
                    "end": 2725.33,
                    "word": " We",
                    "probability": 0.48681640625
                },
                {
                    "start": 2725.33,
                    "end": 2725.95,
                    "word": " reuse",
                    "probability": 0.427734375
                },
                {
                    "start": 2725.95,
                    "end": 2726.17,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 2726.17,
                    "end": 2726.51,
                    "word": " same",
                    "probability": 0.833984375
                },
                {
                    "start": 2726.51,
                    "end": 2727.07,
                    "word": " embedding",
                    "probability": 0.9140625
                },
                {
                    "start": 2727.07,
                    "end": 2727.77,
                    "word": " parameters",
                    "probability": 0.84912109375
                },
                {
                    "start": 2727.77,
                    "end": 2728.01,
                    "word": ",",
                    "probability": 0.443115234375
                },
                {
                    "start": 2728.01,
                    "end": 2728.11,
                    "word": " but",
                    "probability": 0.8974609375
                },
                {
                    "start": 2728.11,
                    "end": 2728.41,
                    "word": " we",
                    "probability": 0.896484375
                },
                {
                    "start": 2728.41,
                    "end": 2728.79,
                    "word": " reduce",
                    "probability": 0.52880859375
                },
                {
                    "start": 2728.79,
                    "end": 2729.37,
                    "word": " the",
                    "probability": 0.8017578125
                },
                {
                    "start": 2729.37,
                    "end": 2730.09,
                    "word": " number",
                    "probability": 0.86181640625
                },
                {
                    "start": 2730.09,
                    "end": 2730.39,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 2730.39,
                    "end": 2731.07,
                    "word": " parameters",
                    "probability": 0.86279296875
                },
                {
                    "start": 2731.07,
                    "end": 2731.55,
                    "word": " for",
                    "probability": 0.7978515625
                }
            ]
        },
        {
            "id": 379,
            "seek": 267033,
            "start": 2731.55,
            "end": 2740.07,
            "text": " transformer blocks by 50%. In total, our model size is 85% of baseline model.",
            "tokens": [
                50696,
                31782,
                8474,
                538,
                2625,
                6856,
                682,
                3217,
                11,
                527,
                2316,
                2744,
                307,
                14695,
                4,
                295,
                20518,
                2316,
                13,
                51096
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2219669137369184,
            "compression_ratio": 1.5494505494505495,
            "no_speech_prob": 0.01506805419921875,
            "words": [
                {
                    "start": 2731.55,
                    "end": 2732.27,
                    "word": " transformer",
                    "probability": 0.037811279296875
                },
                {
                    "start": 2732.27,
                    "end": 2732.71,
                    "word": " blocks",
                    "probability": 0.80810546875
                },
                {
                    "start": 2732.71,
                    "end": 2733.87,
                    "word": " by",
                    "probability": 0.900390625
                },
                {
                    "start": 2733.87,
                    "end": 2734.77,
                    "word": " 50%.",
                    "probability": 0.774658203125
                },
                {
                    "start": 2734.77,
                    "end": 2735.11,
                    "word": " In",
                    "probability": 0.91455078125
                },
                {
                    "start": 2735.11,
                    "end": 2735.55,
                    "word": " total",
                    "probability": 0.77734375
                },
                {
                    "start": 2735.55,
                    "end": 2736.01,
                    "word": ",",
                    "probability": 0.8701171875
                },
                {
                    "start": 2736.01,
                    "end": 2736.27,
                    "word": " our",
                    "probability": 0.81884765625
                },
                {
                    "start": 2736.27,
                    "end": 2736.51,
                    "word": " model",
                    "probability": 0.80810546875
                },
                {
                    "start": 2736.51,
                    "end": 2736.93,
                    "word": " size",
                    "probability": 0.79638671875
                },
                {
                    "start": 2736.93,
                    "end": 2737.21,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 2737.21,
                    "end": 2737.75,
                    "word": " 85",
                    "probability": 0.875
                },
                {
                    "start": 2737.75,
                    "end": 2738.45,
                    "word": "%",
                    "probability": 0.98193359375
                },
                {
                    "start": 2738.45,
                    "end": 2738.77,
                    "word": " of",
                    "probability": 0.896484375
                },
                {
                    "start": 2738.77,
                    "end": 2739.19,
                    "word": " baseline",
                    "probability": 0.75244140625
                },
                {
                    "start": 2739.19,
                    "end": 2739.63,
                    "word": " model",
                    "probability": 0.75390625
                },
                {
                    "start": 2739.63,
                    "end": 2740.07,
                    "word": ".",
                    "probability": 0.9736328125
                }
            ]
        },
        {
            "id": 380,
            "seek": 267033,
            "start": 2740.65,
            "end": 2744.39,
            "text": " When we apply the same mechanism for English version,",
            "tokens": [
                51152,
                1133,
                321,
                3079,
                264,
                912,
                7513,
                337,
                3669,
                3037,
                11,
                51316
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2219669137369184,
            "compression_ratio": 1.5494505494505495,
            "no_speech_prob": 0.01506805419921875,
            "words": [
                {
                    "start": 2740.65,
                    "end": 2741.11,
                    "word": " When",
                    "probability": 0.81005859375
                },
                {
                    "start": 2741.11,
                    "end": 2741.35,
                    "word": " we",
                    "probability": 0.90869140625
                },
                {
                    "start": 2741.35,
                    "end": 2741.69,
                    "word": " apply",
                    "probability": 0.82470703125
                },
                {
                    "start": 2741.69,
                    "end": 2741.89,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 2741.89,
                    "end": 2742.19,
                    "word": " same",
                    "probability": 0.81982421875
                },
                {
                    "start": 2742.19,
                    "end": 2742.63,
                    "word": " mechanism",
                    "probability": 0.409423828125
                },
                {
                    "start": 2742.63,
                    "end": 2743.07,
                    "word": " for",
                    "probability": 0.87548828125
                },
                {
                    "start": 2743.07,
                    "end": 2743.43,
                    "word": " English",
                    "probability": 0.8017578125
                },
                {
                    "start": 2743.43,
                    "end": 2743.95,
                    "word": " version",
                    "probability": 0.54736328125
                },
                {
                    "start": 2743.95,
                    "end": 2744.39,
                    "word": ",",
                    "probability": 0.95361328125
                }
            ]
        },
        {
            "id": 381,
            "seek": 267033,
            "start": 2745.07,
            "end": 2753.77,
            "text": " the English CatBot will have 71% of parameters from baseline model.",
            "tokens": [
                51360,
                264,
                3669,
                9565,
                33,
                310,
                486,
                362,
                30942,
                4,
                295,
                9834,
                490,
                20518,
                2316,
                13,
                51792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2219669137369184,
            "compression_ratio": 1.5494505494505495,
            "no_speech_prob": 0.01506805419921875,
            "words": [
                {
                    "start": 2745.07,
                    "end": 2745.35,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 2745.35,
                    "end": 2746.71,
                    "word": " English",
                    "probability": 0.8193359375
                },
                {
                    "start": 2746.71,
                    "end": 2747.31,
                    "word": " CatBot",
                    "probability": 0.5400899251302084
                },
                {
                    "start": 2747.31,
                    "end": 2747.71,
                    "word": " will",
                    "probability": 0.79296875
                },
                {
                    "start": 2747.71,
                    "end": 2748.07,
                    "word": " have",
                    "probability": 0.86767578125
                },
                {
                    "start": 2748.07,
                    "end": 2749.69,
                    "word": " 71",
                    "probability": 0.89306640625
                },
                {
                    "start": 2749.69,
                    "end": 2750.19,
                    "word": "%",
                    "probability": 0.98974609375
                },
                {
                    "start": 2750.19,
                    "end": 2750.67,
                    "word": " of",
                    "probability": 0.89501953125
                },
                {
                    "start": 2750.67,
                    "end": 2751.33,
                    "word": " parameters",
                    "probability": 0.8583984375
                },
                {
                    "start": 2751.33,
                    "end": 2751.95,
                    "word": " from",
                    "probability": 0.796875
                },
                {
                    "start": 2751.95,
                    "end": 2752.65,
                    "word": " baseline",
                    "probability": 0.8818359375
                },
                {
                    "start": 2752.65,
                    "end": 2753.55,
                    "word": " model",
                    "probability": 0.81396484375
                },
                {
                    "start": 2753.55,
                    "end": 2753.77,
                    "word": ".",
                    "probability": 0.99072265625
                }
            ]
        },
        {
            "id": 382,
            "seek": 269627,
            "start": 2758.49,
            "end": 2765.49,
            "text": " Next, we will inspect how CatBot generates outputs.",
            "tokens": [
                50380,
                3087,
                11,
                321,
                486,
                15018,
                577,
                9565,
                33,
                310,
                23815,
                23930,
                13,
                50660
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2194733727309439,
            "compression_ratio": 1.3825503355704698,
            "no_speech_prob": 0.00872802734375,
            "words": [
                {
                    "start": 2758.49,
                    "end": 2759.23,
                    "word": " Next",
                    "probability": 0.70166015625
                },
                {
                    "start": 2759.23,
                    "end": 2759.47,
                    "word": ",",
                    "probability": 0.814453125
                },
                {
                    "start": 2759.47,
                    "end": 2759.67,
                    "word": " we",
                    "probability": 0.873046875
                },
                {
                    "start": 2759.67,
                    "end": 2759.95,
                    "word": " will",
                    "probability": 0.78125
                },
                {
                    "start": 2759.95,
                    "end": 2760.67,
                    "word": " inspect",
                    "probability": 0.8349609375
                },
                {
                    "start": 2760.67,
                    "end": 2761.35,
                    "word": " how",
                    "probability": 0.869140625
                },
                {
                    "start": 2761.35,
                    "end": 2762.27,
                    "word": " CatBot",
                    "probability": 0.7522786458333334
                },
                {
                    "start": 2762.27,
                    "end": 2763.69,
                    "word": " generates",
                    "probability": 0.7548828125
                },
                {
                    "start": 2763.69,
                    "end": 2764.45,
                    "word": " outputs",
                    "probability": 0.81494140625
                },
                {
                    "start": 2764.45,
                    "end": 2765.49,
                    "word": ".",
                    "probability": 0.931640625
                }
            ]
        },
        {
            "id": 383,
            "seek": 269627,
            "start": 2768.24,
            "end": 2779.28,
            "text": " We use a LIME method to interpret our predictions. LIME is a",
            "tokens": [
                50752,
                492,
                764,
                257,
                441,
                6324,
                36,
                3170,
                281,
                7302,
                527,
                21264,
                13,
                441,
                6324,
                36,
                307,
                257,
                51136
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2194733727309439,
            "compression_ratio": 1.3825503355704698,
            "no_speech_prob": 0.00872802734375,
            "words": [
                {
                    "start": 2768.24,
                    "end": 2768.92,
                    "word": " We",
                    "probability": 0.8642578125
                },
                {
                    "start": 2768.92,
                    "end": 2769.32,
                    "word": " use",
                    "probability": 0.7998046875
                },
                {
                    "start": 2769.94,
                    "end": 2770.96,
                    "word": " a LIME",
                    "probability": 0.9044596354166666
                },
                {
                    "start": 2770.96,
                    "end": 2771.48,
                    "word": " method",
                    "probability": 0.8271484375
                },
                {
                    "start": 2771.48,
                    "end": 2772.04,
                    "word": " to",
                    "probability": 0.8955078125
                },
                {
                    "start": 2772.04,
                    "end": 2772.68,
                    "word": " interpret",
                    "probability": 0.78125
                },
                {
                    "start": 2772.68,
                    "end": 2773.26,
                    "word": " our",
                    "probability": 0.81396484375
                },
                {
                    "start": 2773.26,
                    "end": 2774.36,
                    "word": " predictions",
                    "probability": 0.67822265625
                },
                {
                    "start": 2774.36,
                    "end": 2775.42,
                    "word": ".",
                    "probability": 0.9619140625
                },
                {
                    "start": 2775.42,
                    "end": 2775.82,
                    "word": " LIME",
                    "probability": 0.97412109375
                },
                {
                    "start": 2775.82,
                    "end": 2776.14,
                    "word": " is",
                    "probability": 0.9013671875
                },
                {
                    "start": 2777.08,
                    "end": 2779.28,
                    "word": " a local",
                    "probability": 0.5400390625
                }
            ]
        },
        {
            "id": 384,
            "seek": 269627,
            "start": 2779.28,
            "end": 2786.86,
            "text": " local interpretable model ergodastic explanation method which can be applied to any black box",
            "tokens": [
                51252,
                2654,
                7302,
                712,
                2316,
                1189,
                21787,
                2750,
                10835,
                3170,
                597,
                393,
                312,
                6456,
                281,
                604,
                2211,
                2424,
                51684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2194733727309439,
            "compression_ratio": 1.3825503355704698,
            "no_speech_prob": 0.00872802734375,
            "words": [
                {
                    "start": 2779.28,
                    "end": 2780.58,
                    "word": " interpretable",
                    "probability": 0.77392578125
                },
                {
                    "start": 2780.58,
                    "end": 2781.0,
                    "word": " model",
                    "probability": 0.79931640625
                },
                {
                    "start": 2781.0,
                    "end": 2781.9,
                    "word": " ergodastic",
                    "probability": 0.7303059895833334
                },
                {
                    "start": 2781.9,
                    "end": 2782.7,
                    "word": " explanation",
                    "probability": 0.70068359375
                },
                {
                    "start": 2782.7,
                    "end": 2783.62,
                    "word": " method",
                    "probability": 0.88232421875
                },
                {
                    "start": 2783.62,
                    "end": 2784.38,
                    "word": " which",
                    "probability": 0.52490234375
                },
                {
                    "start": 2784.38,
                    "end": 2784.62,
                    "word": " can",
                    "probability": 0.89501953125
                },
                {
                    "start": 2784.62,
                    "end": 2784.76,
                    "word": " be",
                    "probability": 0.9013671875
                },
                {
                    "start": 2784.76,
                    "end": 2785.12,
                    "word": " applied",
                    "probability": 0.66943359375
                },
                {
                    "start": 2785.12,
                    "end": 2785.38,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 2785.38,
                    "end": 2785.72,
                    "word": " any",
                    "probability": 0.83349609375
                },
                {
                    "start": 2785.72,
                    "end": 2786.54,
                    "word": " black",
                    "probability": 0.759765625
                },
                {
                    "start": 2786.54,
                    "end": 2786.86,
                    "word": " box",
                    "probability": 0.455078125
                }
            ]
        },
        {
            "id": 385,
            "seek": 271733,
            "start": 2786.86,
            "end": 2795.14,
            "text": " models. We can understand a model by perturbing the input and understanding how the predictions",
            "tokens": [
                50384,
                5245,
                13,
                492,
                393,
                1223,
                257,
                2316,
                538,
                280,
                911,
                374,
                4324,
                264,
                4846,
                293,
                3701,
                577,
                264,
                21264,
                50796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23097825439079947,
            "compression_ratio": 1.4122137404580153,
            "no_speech_prob": 0.040985107421875,
            "words": [
                {
                    "start": 2786.86,
                    "end": 2787.82,
                    "word": " models",
                    "probability": 0.311279296875
                },
                {
                    "start": 2787.82,
                    "end": 2788.4,
                    "word": ".",
                    "probability": 0.93505859375
                },
                {
                    "start": 2788.4,
                    "end": 2789.06,
                    "word": " We",
                    "probability": 0.890625
                },
                {
                    "start": 2789.06,
                    "end": 2789.3,
                    "word": " can",
                    "probability": 0.89599609375
                },
                {
                    "start": 2789.3,
                    "end": 2789.82,
                    "word": " understand",
                    "probability": 0.73876953125
                },
                {
                    "start": 2790.14,
                    "end": 2790.4,
                    "word": " a model",
                    "probability": 0.79931640625
                },
                {
                    "start": 2790.4,
                    "end": 2790.78,
                    "word": " by",
                    "probability": 0.90771484375
                },
                {
                    "start": 2790.78,
                    "end": 2791.5,
                    "word": " perturbing",
                    "probability": 0.636322021484375
                },
                {
                    "start": 2791.5,
                    "end": 2791.82,
                    "word": " the",
                    "probability": 0.79638671875
                },
                {
                    "start": 2791.82,
                    "end": 2792.4,
                    "word": " input",
                    "probability": 0.837890625
                },
                {
                    "start": 2792.4,
                    "end": 2793.18,
                    "word": " and",
                    "probability": 0.84033203125
                },
                {
                    "start": 2793.18,
                    "end": 2793.88,
                    "word": " understanding",
                    "probability": 0.66845703125
                },
                {
                    "start": 2793.88,
                    "end": 2794.38,
                    "word": " how",
                    "probability": 0.90380859375
                },
                {
                    "start": 2794.38,
                    "end": 2794.66,
                    "word": " the",
                    "probability": 0.80224609375
                },
                {
                    "start": 2794.66,
                    "end": 2795.14,
                    "word": " predictions",
                    "probability": 0.6240234375
                }
            ]
        },
        {
            "id": 386,
            "seek": 271733,
            "start": 2795.14,
            "end": 2810.25,
            "text": " change. The first LIME example is for a benign email. The prediction score for this email",
            "tokens": [
                50796,
                1319,
                13,
                440,
                700,
                441,
                6324,
                36,
                1365,
                307,
                337,
                257,
                3271,
                788,
                3796,
                13,
                440,
                17630,
                6175,
                337,
                341,
                3796,
                51412
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23097825439079947,
            "compression_ratio": 1.4122137404580153,
            "no_speech_prob": 0.040985107421875,
            "words": [
                {
                    "start": 2795.14,
                    "end": 2796.02,
                    "word": " change",
                    "probability": 0.86328125
                },
                {
                    "start": 2796.02,
                    "end": 2797.1,
                    "word": ".",
                    "probability": 0.974609375
                },
                {
                    "start": 2799.43,
                    "end": 2800.37,
                    "word": " The",
                    "probability": 0.80126953125
                },
                {
                    "start": 2800.37,
                    "end": 2800.87,
                    "word": " first",
                    "probability": 0.806640625
                },
                {
                    "start": 2800.87,
                    "end": 2801.39,
                    "word": " LIME",
                    "probability": 0.8123372395833334
                },
                {
                    "start": 2801.39,
                    "end": 2802.01,
                    "word": " example",
                    "probability": 0.92333984375
                },
                {
                    "start": 2802.01,
                    "end": 2802.45,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 2802.45,
                    "end": 2802.97,
                    "word": " for",
                    "probability": 0.77490234375
                },
                {
                    "start": 2803.55,
                    "end": 2804.67,
                    "word": " a benign",
                    "probability": 0.896728515625
                },
                {
                    "start": 2804.67,
                    "end": 2805.13,
                    "word": " email",
                    "probability": 0.8134765625
                },
                {
                    "start": 2805.13,
                    "end": 2806.21,
                    "word": ".",
                    "probability": 0.94580078125
                },
                {
                    "start": 2806.89,
                    "end": 2807.47,
                    "word": " The",
                    "probability": 0.796875
                },
                {
                    "start": 2807.47,
                    "end": 2807.97,
                    "word": " prediction",
                    "probability": 0.85791015625
                },
                {
                    "start": 2807.97,
                    "end": 2808.69,
                    "word": " score",
                    "probability": 0.828125
                },
                {
                    "start": 2808.69,
                    "end": 2809.21,
                    "word": " for",
                    "probability": 0.86962890625
                },
                {
                    "start": 2809.21,
                    "end": 2809.77,
                    "word": " this",
                    "probability": 0.87451171875
                },
                {
                    "start": 2809.77,
                    "end": 2810.25,
                    "word": " email",
                    "probability": 0.89111328125
                }
            ]
        },
        {
            "id": 387,
            "seek": 274283,
            "start": 2811.01,
            "end": 2817.53,
            "text": " is close to zero. And we highlight legitimate tokens with",
            "tokens": [
                50412,
                307,
                1998,
                281,
                4018,
                13,
                400,
                321,
                5078,
                17956,
                22667,
                365,
                50712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.29498921180593557,
            "compression_ratio": 1.4743589743589745,
            "no_speech_prob": 0.267578125,
            "words": [
                {
                    "start": 2811.01,
                    "end": 2811.45,
                    "word": " is",
                    "probability": 0.06951904296875
                },
                {
                    "start": 2811.45,
                    "end": 2811.91,
                    "word": " close",
                    "probability": 0.162353515625
                },
                {
                    "start": 2811.91,
                    "end": 2812.55,
                    "word": " to",
                    "probability": 0.7001953125
                },
                {
                    "start": 2812.55,
                    "end": 2812.87,
                    "word": " zero",
                    "probability": 0.55224609375
                },
                {
                    "start": 2812.87,
                    "end": 2813.75,
                    "word": ".",
                    "probability": 0.427490234375
                },
                {
                    "start": 2813.79,
                    "end": 2814.37,
                    "word": " And",
                    "probability": 0.591796875
                },
                {
                    "start": 2814.37,
                    "end": 2814.63,
                    "word": " we",
                    "probability": 0.85986328125
                },
                {
                    "start": 2814.63,
                    "end": 2815.27,
                    "word": " highlight",
                    "probability": 0.7744140625
                },
                {
                    "start": 2815.27,
                    "end": 2816.23,
                    "word": " legitimate",
                    "probability": 0.70556640625
                },
                {
                    "start": 2816.23,
                    "end": 2816.85,
                    "word": " tokens",
                    "probability": 0.75537109375
                },
                {
                    "start": 2816.85,
                    "end": 2817.53,
                    "word": " with",
                    "probability": 0.80712890625
                }
            ]
        },
        {
            "id": 388,
            "seek": 274283,
            "start": 2819.37,
            "end": 2826.19,
            "text": " blue color and malicious ones as orange one. And this one, we don't have any high-rated",
            "tokens": [
                50824,
                3344,
                2017,
                293,
                33496,
                2306,
                382,
                7671,
                472,
                13,
                400,
                341,
                472,
                11,
                321,
                500,
                380,
                362,
                604,
                1090,
                12,
                5468,
                51172
            ],
            "temperature": 0.0,
            "avg_logprob": -0.29498921180593557,
            "compression_ratio": 1.4743589743589745,
            "no_speech_prob": 0.267578125,
            "words": [
                {
                    "start": 2819.37,
                    "end": 2819.81,
                    "word": " blue",
                    "probability": 0.87451171875
                },
                {
                    "start": 2819.81,
                    "end": 2820.23,
                    "word": " color",
                    "probability": 0.74853515625
                },
                {
                    "start": 2820.23,
                    "end": 2820.67,
                    "word": " and",
                    "probability": 0.841796875
                },
                {
                    "start": 2820.67,
                    "end": 2821.01,
                    "word": " malicious",
                    "probability": 0.7470703125
                },
                {
                    "start": 2821.01,
                    "end": 2821.57,
                    "word": " ones",
                    "probability": 0.8359375
                },
                {
                    "start": 2821.57,
                    "end": 2821.89,
                    "word": " as",
                    "probability": 0.75341796875
                },
                {
                    "start": 2821.89,
                    "end": 2822.37,
                    "word": " orange",
                    "probability": 0.62646484375
                },
                {
                    "start": 2822.37,
                    "end": 2822.73,
                    "word": " one",
                    "probability": 0.51513671875
                },
                {
                    "start": 2822.73,
                    "end": 2823.25,
                    "word": ".",
                    "probability": 0.8525390625
                },
                {
                    "start": 2823.25,
                    "end": 2823.61,
                    "word": " And",
                    "probability": 0.82763671875
                },
                {
                    "start": 2823.61,
                    "end": 2823.87,
                    "word": " this",
                    "probability": 0.84716796875
                },
                {
                    "start": 2823.87,
                    "end": 2824.15,
                    "word": " one",
                    "probability": 0.81787109375
                },
                {
                    "start": 2824.15,
                    "end": 2824.31,
                    "word": ",",
                    "probability": 0.59326171875
                },
                {
                    "start": 2824.31,
                    "end": 2824.37,
                    "word": " we",
                    "probability": 0.90478515625
                },
                {
                    "start": 2824.37,
                    "end": 2824.55,
                    "word": " don't",
                    "probability": 0.94287109375
                },
                {
                    "start": 2824.55,
                    "end": 2824.73,
                    "word": " have",
                    "probability": 0.88623046875
                },
                {
                    "start": 2824.73,
                    "end": 2825.07,
                    "word": " any",
                    "probability": 0.81689453125
                },
                {
                    "start": 2825.07,
                    "end": 2825.49,
                    "word": " high",
                    "probability": 0.84521484375
                },
                {
                    "start": 2825.49,
                    "end": 2826.19,
                    "word": "-rated",
                    "probability": 0.41400146484375
                }
            ]
        },
        {
            "id": 389,
            "seek": 274283,
            "start": 2826.19,
            "end": 2837.5,
            "text": " tokens for malicious ones. Next, we have a BC sample. The model prediction score for",
            "tokens": [
                51172,
                22667,
                337,
                33496,
                2306,
                13,
                3087,
                11,
                321,
                362,
                257,
                14359,
                6889,
                13,
                440,
                2316,
                17630,
                6175,
                337,
                51632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.29498921180593557,
            "compression_ratio": 1.4743589743589745,
            "no_speech_prob": 0.267578125,
            "words": [
                {
                    "start": 2826.19,
                    "end": 2826.89,
                    "word": " tokens",
                    "probability": 0.73291015625
                },
                {
                    "start": 2826.89,
                    "end": 2827.33,
                    "word": " for",
                    "probability": 0.86181640625
                },
                {
                    "start": 2827.33,
                    "end": 2827.77,
                    "word": " malicious",
                    "probability": 0.833984375
                },
                {
                    "start": 2827.77,
                    "end": 2828.75,
                    "word": " ones",
                    "probability": 0.82958984375
                },
                {
                    "start": 2828.75,
                    "end": 2829.63,
                    "word": ".",
                    "probability": 0.9892578125
                },
                {
                    "start": 2831.82,
                    "end": 2832.7,
                    "word": " Next",
                    "probability": 0.8818359375
                },
                {
                    "start": 2832.7,
                    "end": 2832.94,
                    "word": ",",
                    "probability": 0.740234375
                },
                {
                    "start": 2832.94,
                    "end": 2833.1,
                    "word": " we",
                    "probability": 0.904296875
                },
                {
                    "start": 2833.1,
                    "end": 2833.44,
                    "word": " have",
                    "probability": 0.86572265625
                },
                {
                    "start": 2833.82,
                    "end": 2834.1,
                    "word": " a BC",
                    "probability": 0.37744140625
                },
                {
                    "start": 2834.1,
                    "end": 2834.58,
                    "word": " sample",
                    "probability": 0.7041015625
                },
                {
                    "start": 2834.58,
                    "end": 2835.46,
                    "word": ".",
                    "probability": 0.9541015625
                },
                {
                    "start": 2835.62,
                    "end": 2835.88,
                    "word": " The",
                    "probability": 0.8115234375
                },
                {
                    "start": 2835.88,
                    "end": 2836.2,
                    "word": " model",
                    "probability": 0.73388671875
                },
                {
                    "start": 2836.2,
                    "end": 2836.68,
                    "word": " prediction",
                    "probability": 0.5732421875
                },
                {
                    "start": 2836.68,
                    "end": 2837.2,
                    "word": " score",
                    "probability": 0.78369140625
                },
                {
                    "start": 2837.2,
                    "end": 2837.5,
                    "word": " for",
                    "probability": 0.85693359375
                }
            ]
        },
        {
            "id": 390,
            "seek": 277283,
            "start": 2837.5,
            "end": 2848.86,
            "text": " maliciousness is close to one. And the model recognizes transfer and urgent payment as",
            "tokens": [
                50364,
                33496,
                1287,
                307,
                1998,
                281,
                472,
                13,
                400,
                264,
                2316,
                26564,
                5003,
                293,
                19022,
                10224,
                382,
                50944
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3599137967002803,
            "compression_ratio": 1.535031847133758,
            "no_speech_prob": 0.042236328125,
            "words": [
                {
                    "start": 2837.5,
                    "end": 2838.24,
                    "word": " maliciousness",
                    "probability": 0.559814453125
                },
                {
                    "start": 2838.24,
                    "end": 2838.64,
                    "word": " is",
                    "probability": 0.875
                },
                {
                    "start": 2838.64,
                    "end": 2838.98,
                    "word": " close",
                    "probability": 0.06182861328125
                },
                {
                    "start": 2838.98,
                    "end": 2839.18,
                    "word": " to",
                    "probability": 0.68603515625
                },
                {
                    "start": 2839.18,
                    "end": 2839.52,
                    "word": " one",
                    "probability": 0.2034912109375
                },
                {
                    "start": 2839.52,
                    "end": 2840.42,
                    "word": ".",
                    "probability": 0.32763671875
                },
                {
                    "start": 2840.42,
                    "end": 2840.86,
                    "word": " And",
                    "probability": 0.7197265625
                },
                {
                    "start": 2840.86,
                    "end": 2841.34,
                    "word": " the",
                    "probability": 0.7939453125
                },
                {
                    "start": 2841.34,
                    "end": 2841.6,
                    "word": " model",
                    "probability": 0.8173828125
                },
                {
                    "start": 2841.6,
                    "end": 2843.12,
                    "word": " recognizes",
                    "probability": 0.353515625
                },
                {
                    "start": 2843.12,
                    "end": 2845.1,
                    "word": " transfer",
                    "probability": 0.474609375
                },
                {
                    "start": 2845.1,
                    "end": 2846.12,
                    "word": " and",
                    "probability": 0.8076171875
                },
                {
                    "start": 2846.12,
                    "end": 2846.88,
                    "word": " urgent",
                    "probability": 0.4814453125
                },
                {
                    "start": 2846.88,
                    "end": 2847.7,
                    "word": " payment",
                    "probability": 0.8310546875
                },
                {
                    "start": 2847.7,
                    "end": 2848.86,
                    "word": " as",
                    "probability": 0.888671875
                }
            ]
        },
        {
            "id": 391,
            "seek": 277283,
            "start": 2848.86,
            "end": 2862.17,
            "text": " high-rated tokens. Next, we have another BC sample, which is related with gift card.",
            "tokens": [
                50944,
                1090,
                12,
                5468,
                22667,
                13,
                3087,
                11,
                321,
                362,
                1071,
                14359,
                6889,
                11,
                597,
                307,
                4077,
                365,
                5306,
                2920,
                13,
                51432
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3599137967002803,
            "compression_ratio": 1.535031847133758,
            "no_speech_prob": 0.042236328125,
            "words": [
                {
                    "start": 2848.86,
                    "end": 2849.5,
                    "word": " high",
                    "probability": 0.837890625
                },
                {
                    "start": 2849.5,
                    "end": 2849.88,
                    "word": "-rated",
                    "probability": 0.368927001953125
                },
                {
                    "start": 2849.88,
                    "end": 2850.48,
                    "word": " tokens",
                    "probability": 0.763671875
                },
                {
                    "start": 2850.48,
                    "end": 2851.44,
                    "word": ".",
                    "probability": 0.990234375
                },
                {
                    "start": 2854.73,
                    "end": 2855.49,
                    "word": " Next",
                    "probability": 0.884765625
                },
                {
                    "start": 2855.49,
                    "end": 2855.65,
                    "word": ",",
                    "probability": 0.79638671875
                },
                {
                    "start": 2855.65,
                    "end": 2855.77,
                    "word": " we",
                    "probability": 0.9052734375
                },
                {
                    "start": 2855.77,
                    "end": 2855.97,
                    "word": " have",
                    "probability": 0.87744140625
                },
                {
                    "start": 2855.97,
                    "end": 2856.41,
                    "word": " another",
                    "probability": 0.8271484375
                },
                {
                    "start": 2856.41,
                    "end": 2856.77,
                    "word": " BC",
                    "probability": 0.0013942718505859375
                },
                {
                    "start": 2856.77,
                    "end": 2857.29,
                    "word": " sample",
                    "probability": 0.6025390625
                },
                {
                    "start": 2857.29,
                    "end": 2858.35,
                    "word": ",",
                    "probability": 0.62744140625
                },
                {
                    "start": 2858.35,
                    "end": 2858.81,
                    "word": " which",
                    "probability": 0.921875
                },
                {
                    "start": 2858.81,
                    "end": 2859.01,
                    "word": " is",
                    "probability": 0.896484375
                },
                {
                    "start": 2859.01,
                    "end": 2859.49,
                    "word": " related",
                    "probability": 0.6796875
                },
                {
                    "start": 2859.49,
                    "end": 2860.37,
                    "word": " with",
                    "probability": 0.76953125
                },
                {
                    "start": 2860.37,
                    "end": 2861.09,
                    "word": " gift",
                    "probability": 0.77587890625
                },
                {
                    "start": 2861.09,
                    "end": 2861.55,
                    "word": " card",
                    "probability": 0.7314453125
                },
                {
                    "start": 2861.55,
                    "end": 2862.17,
                    "word": ".",
                    "probability": 0.77587890625
                }
            ]
        },
        {
            "id": 392,
            "seek": 277283,
            "start": 2862.17,
            "end": 2868.97,
            "text": " And the model prediction score is close to one and card and urgently.",
            "tokens": [
                51464,
                400,
                264,
                2316,
                17630,
                6175,
                307,
                1998,
                281,
                472,
                293,
                2920,
                293,
                19022,
                356,
                13,
                51780
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3599137967002803,
            "compression_ratio": 1.535031847133758,
            "no_speech_prob": 0.042236328125,
            "words": [
                {
                    "start": 2862.17,
                    "end": 2862.51,
                    "word": " And",
                    "probability": 0.87548828125
                },
                {
                    "start": 2862.51,
                    "end": 2862.67,
                    "word": " the",
                    "probability": 0.7978515625
                },
                {
                    "start": 2862.67,
                    "end": 2862.87,
                    "word": " model",
                    "probability": 0.8017578125
                },
                {
                    "start": 2862.87,
                    "end": 2863.33,
                    "word": " prediction",
                    "probability": 0.80224609375
                },
                {
                    "start": 2863.33,
                    "end": 2863.83,
                    "word": " score",
                    "probability": 0.51611328125
                },
                {
                    "start": 2863.83,
                    "end": 2864.29,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 2864.29,
                    "end": 2864.61,
                    "word": " close",
                    "probability": 0.76953125
                },
                {
                    "start": 2864.61,
                    "end": 2864.77,
                    "word": " to",
                    "probability": 0.88037109375
                },
                {
                    "start": 2864.77,
                    "end": 2865.19,
                    "word": " one",
                    "probability": 0.7685546875
                },
                {
                    "start": 2865.19,
                    "end": 2865.91,
                    "word": " and",
                    "probability": 0.360595703125
                },
                {
                    "start": 2865.91,
                    "end": 2866.91,
                    "word": " card",
                    "probability": 0.779296875
                },
                {
                    "start": 2866.91,
                    "end": 2867.41,
                    "word": " and",
                    "probability": 0.63525390625
                },
                {
                    "start": 2867.41,
                    "end": 2868.45,
                    "word": " urgently",
                    "probability": 0.628173828125
                },
                {
                    "start": 2868.45,
                    "end": 2868.97,
                    "word": ".",
                    "probability": 0.98779296875
                }
            ]
        },
        {
            "id": 393,
            "seek": 279277,
            "start": 2870.11,
            "end": 2883.71,
            "text": " The tokens are high-rated for this email. Next, we have two handcrafted social engineering emails.",
            "tokens": [
                50380,
                440,
                22667,
                366,
                1090,
                12,
                5468,
                337,
                341,
                3796,
                13,
                3087,
                11,
                321,
                362,
                732,
                1011,
                5611,
                292,
                2093,
                7043,
                12524,
                13,
                50920
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24728954081632654,
            "compression_ratio": 1.3673469387755102,
            "no_speech_prob": 0.0060272216796875,
            "words": [
                {
                    "start": 2870.11,
                    "end": 2870.71,
                    "word": " The",
                    "probability": 0.127197265625
                },
                {
                    "start": 2870.71,
                    "end": 2871.15,
                    "word": " tokens",
                    "probability": 0.56494140625
                },
                {
                    "start": 2871.15,
                    "end": 2871.49,
                    "word": " are",
                    "probability": 0.87451171875
                },
                {
                    "start": 2871.49,
                    "end": 2871.83,
                    "word": " high",
                    "probability": 0.7294921875
                },
                {
                    "start": 2871.83,
                    "end": 2872.27,
                    "word": "-rated",
                    "probability": 0.32098388671875
                },
                {
                    "start": 2872.27,
                    "end": 2872.91,
                    "word": " for",
                    "probability": 0.84375
                },
                {
                    "start": 2872.91,
                    "end": 2873.31,
                    "word": " this",
                    "probability": 0.7958984375
                },
                {
                    "start": 2873.31,
                    "end": 2873.75,
                    "word": " email",
                    "probability": 0.87158203125
                },
                {
                    "start": 2873.75,
                    "end": 2874.59,
                    "word": ".",
                    "probability": 0.548828125
                },
                {
                    "start": 2878.07,
                    "end": 2878.43,
                    "word": " Next",
                    "probability": 0.853515625
                },
                {
                    "start": 2878.43,
                    "end": 2878.63,
                    "word": ",",
                    "probability": 0.455322265625
                },
                {
                    "start": 2878.63,
                    "end": 2878.69,
                    "word": " we",
                    "probability": 0.890625
                },
                {
                    "start": 2878.69,
                    "end": 2878.93,
                    "word": " have",
                    "probability": 0.87841796875
                },
                {
                    "start": 2878.93,
                    "end": 2879.45,
                    "word": " two",
                    "probability": 0.7470703125
                },
                {
                    "start": 2879.45,
                    "end": 2881.85,
                    "word": " handcrafted",
                    "probability": 0.72119140625
                },
                {
                    "start": 2881.85,
                    "end": 2882.27,
                    "word": " social",
                    "probability": 0.61181640625
                },
                {
                    "start": 2882.27,
                    "end": 2882.85,
                    "word": " engineering",
                    "probability": 0.79541015625
                },
                {
                    "start": 2882.85,
                    "end": 2883.33,
                    "word": " emails",
                    "probability": 0.75830078125
                },
                {
                    "start": 2883.33,
                    "end": 2883.71,
                    "word": ".",
                    "probability": 0.9189453125
                }
            ]
        },
        {
            "id": 394,
            "seek": 279277,
            "start": 2883.71,
            "end": 2892.57,
            "text": " They look quite different, but if you read the text carefully, they are asking the same wire transfer.",
            "tokens": [
                50920,
                814,
                574,
                1596,
                819,
                11,
                457,
                498,
                291,
                1401,
                264,
                2487,
                7500,
                11,
                436,
                366,
                3365,
                264,
                912,
                6234,
                5003,
                13,
                51344
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24728954081632654,
            "compression_ratio": 1.3673469387755102,
            "no_speech_prob": 0.0060272216796875,
            "words": [
                {
                    "start": 2883.71,
                    "end": 2884.05,
                    "word": " They",
                    "probability": 0.66748046875
                },
                {
                    "start": 2884.05,
                    "end": 2884.41,
                    "word": " look",
                    "probability": 0.90576171875
                },
                {
                    "start": 2884.41,
                    "end": 2884.75,
                    "word": " quite",
                    "probability": 0.8759765625
                },
                {
                    "start": 2884.75,
                    "end": 2885.33,
                    "word": " different",
                    "probability": 0.69873046875
                },
                {
                    "start": 2885.33,
                    "end": 2885.55,
                    "word": ",",
                    "probability": 0.62890625
                },
                {
                    "start": 2885.55,
                    "end": 2885.67,
                    "word": " but",
                    "probability": 0.88720703125
                },
                {
                    "start": 2885.67,
                    "end": 2885.93,
                    "word": " if",
                    "probability": 0.8388671875
                },
                {
                    "start": 2885.93,
                    "end": 2886.15,
                    "word": " you",
                    "probability": 0.75048828125
                },
                {
                    "start": 2886.15,
                    "end": 2886.67,
                    "word": " read",
                    "probability": 0.87646484375
                },
                {
                    "start": 2886.67,
                    "end": 2887.09,
                    "word": " the",
                    "probability": 0.7939453125
                },
                {
                    "start": 2887.09,
                    "end": 2887.37,
                    "word": " text",
                    "probability": 0.85693359375
                },
                {
                    "start": 2887.37,
                    "end": 2887.97,
                    "word": " carefully",
                    "probability": 0.7060546875
                },
                {
                    "start": 2887.97,
                    "end": 2888.71,
                    "word": ",",
                    "probability": 0.8349609375
                },
                {
                    "start": 2888.71,
                    "end": 2888.91,
                    "word": " they",
                    "probability": 0.7197265625
                },
                {
                    "start": 2888.91,
                    "end": 2889.11,
                    "word": " are",
                    "probability": 0.69970703125
                },
                {
                    "start": 2889.11,
                    "end": 2889.71,
                    "word": " asking",
                    "probability": 0.810546875
                },
                {
                    "start": 2889.71,
                    "end": 2890.61,
                    "word": " the",
                    "probability": 0.77880859375
                },
                {
                    "start": 2890.61,
                    "end": 2891.03,
                    "word": " same",
                    "probability": 0.826171875
                },
                {
                    "start": 2891.03,
                    "end": 2891.63,
                    "word": " wire",
                    "probability": 0.6318359375
                },
                {
                    "start": 2891.63,
                    "end": 2892.15,
                    "word": " transfer",
                    "probability": 0.89990234375
                },
                {
                    "start": 2892.15,
                    "end": 2892.57,
                    "word": ".",
                    "probability": 0.8544921875
                }
            ]
        },
        {
            "id": 395,
            "seek": 281667,
            "start": 2893.45,
            "end": 2902.85,
            "text": " And the model predictions are close to one for both emails and the highlight tokens are payment",
            "tokens": [
                50412,
                400,
                264,
                2316,
                21264,
                366,
                1998,
                281,
                472,
                337,
                1293,
                12524,
                293,
                264,
                5078,
                22667,
                366,
                10224,
                50880
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3533653723887908,
            "compression_ratio": 1.3740458015267176,
            "no_speech_prob": 0.13525390625,
            "words": [
                {
                    "start": 2893.45,
                    "end": 2894.05,
                    "word": " And",
                    "probability": 0.036834716796875
                },
                {
                    "start": 2894.05,
                    "end": 2895.85,
                    "word": " the",
                    "probability": 0.11181640625
                },
                {
                    "start": 2895.85,
                    "end": 2896.11,
                    "word": " model",
                    "probability": 0.79345703125
                },
                {
                    "start": 2896.11,
                    "end": 2897.11,
                    "word": " predictions",
                    "probability": 0.360107421875
                },
                {
                    "start": 2897.11,
                    "end": 2897.55,
                    "word": " are",
                    "probability": 0.44677734375
                },
                {
                    "start": 2897.55,
                    "end": 2897.85,
                    "word": " close",
                    "probability": 0.051971435546875
                },
                {
                    "start": 2897.85,
                    "end": 2898.01,
                    "word": " to",
                    "probability": 0.1165771484375
                },
                {
                    "start": 2898.01,
                    "end": 2898.19,
                    "word": " one",
                    "probability": 0.4287109375
                },
                {
                    "start": 2898.19,
                    "end": 2898.55,
                    "word": " for",
                    "probability": 0.822265625
                },
                {
                    "start": 2898.55,
                    "end": 2899.03,
                    "word": " both",
                    "probability": 0.82861328125
                },
                {
                    "start": 2899.03,
                    "end": 2899.63,
                    "word": " emails",
                    "probability": 0.62451171875
                },
                {
                    "start": 2899.63,
                    "end": 2900.25,
                    "word": " and",
                    "probability": 0.515625
                },
                {
                    "start": 2900.25,
                    "end": 2900.97,
                    "word": " the",
                    "probability": 0.70947265625
                },
                {
                    "start": 2900.97,
                    "end": 2901.53,
                    "word": " highlight",
                    "probability": 0.59326171875
                },
                {
                    "start": 2901.53,
                    "end": 2902.15,
                    "word": " tokens",
                    "probability": 0.65625
                },
                {
                    "start": 2902.15,
                    "end": 2902.45,
                    "word": " are",
                    "probability": 0.81005859375
                },
                {
                    "start": 2902.45,
                    "end": 2902.85,
                    "word": " payment",
                    "probability": 0.85205078125
                }
            ]
        },
        {
            "id": 396,
            "seek": 281667,
            "start": 2904.05,
            "end": 2916.47,
            "text": " and swift or as soon as possible. These examples demonstrated our model's ability to",
            "tokens": [
                50940,
                293,
                29184,
                420,
                382,
                2321,
                382,
                1944,
                13,
                1981,
                5110,
                18772,
                527,
                2316,
                311,
                3485,
                281,
                51544
            ],
            "temperature": 0.0,
            "avg_logprob": -0.3533653723887908,
            "compression_ratio": 1.3740458015267176,
            "no_speech_prob": 0.13525390625,
            "words": [
                {
                    "start": 2904.05,
                    "end": 2904.65,
                    "word": " and",
                    "probability": 0.7490234375
                },
                {
                    "start": 2904.65,
                    "end": 2906.61,
                    "word": " swift",
                    "probability": 0.1400146484375
                },
                {
                    "start": 2906.61,
                    "end": 2907.87,
                    "word": " or",
                    "probability": 0.5517578125
                },
                {
                    "start": 2907.87,
                    "end": 2908.39,
                    "word": " as",
                    "probability": 0.8642578125
                },
                {
                    "start": 2908.39,
                    "end": 2908.65,
                    "word": " soon",
                    "probability": 0.87255859375
                },
                {
                    "start": 2908.65,
                    "end": 2908.93,
                    "word": " as",
                    "probability": 0.9033203125
                },
                {
                    "start": 2908.93,
                    "end": 2909.53,
                    "word": " possible",
                    "probability": 0.83447265625
                },
                {
                    "start": 2909.53,
                    "end": 2910.73,
                    "word": ".",
                    "probability": 0.92822265625
                },
                {
                    "start": 2911.03,
                    "end": 2911.47,
                    "word": " These",
                    "probability": 0.65869140625
                },
                {
                    "start": 2911.47,
                    "end": 2912.09,
                    "word": " examples",
                    "probability": 0.7412109375
                },
                {
                    "start": 2912.09,
                    "end": 2912.99,
                    "word": " demonstrated",
                    "probability": 0.449951171875
                },
                {
                    "start": 2912.99,
                    "end": 2913.93,
                    "word": " our",
                    "probability": 0.8291015625
                },
                {
                    "start": 2913.93,
                    "end": 2915.11,
                    "word": " model's",
                    "probability": 0.787109375
                },
                {
                    "start": 2915.11,
                    "end": 2915.69,
                    "word": " ability",
                    "probability": 0.849609375
                },
                {
                    "start": 2915.69,
                    "end": 2916.47,
                    "word": " to",
                    "probability": 0.9111328125
                }
            ]
        },
        {
            "id": 397,
            "seek": 284163,
            "start": 2917.19,
            "end": 2925.39,
            "text": " understand complex texts and conceptually similar emails can be identified.",
            "tokens": [
                50400,
                1223,
                3997,
                15765,
                293,
                3410,
                671,
                2531,
                12524,
                393,
                312,
                9234,
                13,
                50788
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27613146654490767,
            "compression_ratio": 1.4337349397590362,
            "no_speech_prob": 0.0188140869140625,
            "words": [
                {
                    "start": 2917.19,
                    "end": 2917.91,
                    "word": " understand",
                    "probability": 0.1417236328125
                },
                {
                    "start": 2917.91,
                    "end": 2919.31,
                    "word": " complex",
                    "probability": 0.7587890625
                },
                {
                    "start": 2919.31,
                    "end": 2919.83,
                    "word": " texts",
                    "probability": 0.397705078125
                },
                {
                    "start": 2919.83,
                    "end": 2920.91,
                    "word": " and",
                    "probability": 0.6484375
                },
                {
                    "start": 2920.91,
                    "end": 2922.19,
                    "word": " conceptually",
                    "probability": 0.762939453125
                },
                {
                    "start": 2922.19,
                    "end": 2922.89,
                    "word": " similar",
                    "probability": 0.89990234375
                },
                {
                    "start": 2922.89,
                    "end": 2923.65,
                    "word": " emails",
                    "probability": 0.08172607421875
                },
                {
                    "start": 2923.65,
                    "end": 2923.97,
                    "word": " can",
                    "probability": 0.86962890625
                },
                {
                    "start": 2923.97,
                    "end": 2924.17,
                    "word": " be",
                    "probability": 0.90625
                },
                {
                    "start": 2924.17,
                    "end": 2924.81,
                    "word": " identified",
                    "probability": 0.88330078125
                },
                {
                    "start": 2924.81,
                    "end": 2925.39,
                    "word": ".",
                    "probability": 0.81787109375
                }
            ]
        },
        {
            "id": 398,
            "seek": 284163,
            "start": 2928.83,
            "end": 2937.15,
            "text": " In conclusion, our CatBot is a carefully re-architectural transform-based model.",
            "tokens": [
                50880,
                682,
                10063,
                11,
                527,
                9565,
                33,
                310,
                307,
                257,
                7500,
                319,
                12,
                1178,
                5739,
                1807,
                4088,
                12,
                6032,
                2316,
                13,
                51264
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27613146654490767,
            "compression_ratio": 1.4337349397590362,
            "no_speech_prob": 0.0188140869140625,
            "words": [
                {
                    "start": 2928.83,
                    "end": 2929.65,
                    "word": " In",
                    "probability": 0.87353515625
                },
                {
                    "start": 2929.65,
                    "end": 2930.21,
                    "word": " conclusion",
                    "probability": 0.8056640625
                },
                {
                    "start": 2930.21,
                    "end": 2932.03,
                    "word": ",",
                    "probability": 0.904296875
                },
                {
                    "start": 2932.03,
                    "end": 2932.23,
                    "word": " our",
                    "probability": 0.595703125
                },
                {
                    "start": 2932.23,
                    "end": 2932.85,
                    "word": " CatBot",
                    "probability": 0.6502278645833334
                },
                {
                    "start": 2932.85,
                    "end": 2933.25,
                    "word": " is",
                    "probability": 0.9091796875
                },
                {
                    "start": 2933.63,
                    "end": 2934.11,
                    "word": " a carefully",
                    "probability": 0.6875
                },
                {
                    "start": 2934.11,
                    "end": 2934.49,
                    "word": " re",
                    "probability": 0.84423828125
                },
                {
                    "start": 2934.49,
                    "end": 2935.23,
                    "word": "-architectural",
                    "probability": 0.7357177734375
                },
                {
                    "start": 2935.23,
                    "end": 2935.83,
                    "word": " transform",
                    "probability": 0.71875
                },
                {
                    "start": 2935.83,
                    "end": 2936.19,
                    "word": "-based",
                    "probability": 0.741943359375
                },
                {
                    "start": 2936.19,
                    "end": 2936.53,
                    "word": " model",
                    "probability": 0.66357421875
                },
                {
                    "start": 2936.53,
                    "end": 2937.15,
                    "word": ".",
                    "probability": 0.91796875
                }
            ]
        },
        {
            "id": 399,
            "seek": 284163,
            "start": 2937.15,
            "end": 2943.95,
            "text": " With this architecture, we achieve both high-speed and high-accuracy in detecting",
            "tokens": [
                51264,
                2022,
                341,
                9482,
                11,
                321,
                4584,
                1293,
                1090,
                12,
                22746,
                293,
                1090,
                12,
                8476,
                374,
                2551,
                294,
                40237,
                51612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27613146654490767,
            "compression_ratio": 1.4337349397590362,
            "no_speech_prob": 0.0188140869140625,
            "words": [
                {
                    "start": 2937.15,
                    "end": 2937.33,
                    "word": " With",
                    "probability": 0.87158203125
                },
                {
                    "start": 2937.33,
                    "end": 2938.03,
                    "word": " this",
                    "probability": 0.88525390625
                },
                {
                    "start": 2938.03,
                    "end": 2938.87,
                    "word": " architecture",
                    "probability": 0.8134765625
                },
                {
                    "start": 2938.87,
                    "end": 2939.05,
                    "word": ",",
                    "probability": 0.74609375
                },
                {
                    "start": 2939.05,
                    "end": 2939.21,
                    "word": " we",
                    "probability": 0.89892578125
                },
                {
                    "start": 2939.21,
                    "end": 2939.79,
                    "word": " achieve",
                    "probability": 0.61083984375
                },
                {
                    "start": 2939.79,
                    "end": 2940.39,
                    "word": " both",
                    "probability": 0.6884765625
                },
                {
                    "start": 2940.39,
                    "end": 2941.11,
                    "word": " high",
                    "probability": 0.85546875
                },
                {
                    "start": 2941.11,
                    "end": 2941.47,
                    "word": "-speed",
                    "probability": 0.689697265625
                },
                {
                    "start": 2941.47,
                    "end": 2942.05,
                    "word": " and",
                    "probability": 0.88134765625
                },
                {
                    "start": 2942.05,
                    "end": 2942.49,
                    "word": " high",
                    "probability": 0.85791015625
                },
                {
                    "start": 2942.49,
                    "end": 2943.13,
                    "word": "-accuracy",
                    "probability": 0.9266357421875
                },
                {
                    "start": 2943.13,
                    "end": 2943.45,
                    "word": " in",
                    "probability": 0.87255859375
                },
                {
                    "start": 2943.45,
                    "end": 2943.95,
                    "word": " detecting",
                    "probability": 0.828125
                }
            ]
        },
        {
            "id": 400,
            "seek": 286877,
            "start": 2945.01,
            "end": 2951.29,
            "text": " handicrafted social engineering email attacks. In the future, we want to apply the same",
            "tokens": [
                50392,
                31369,
                4469,
                292,
                2093,
                7043,
                3796,
                8122,
                13,
                682,
                264,
                2027,
                11,
                321,
                528,
                281,
                3079,
                264,
                912,
                50744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23833333094914755,
            "compression_ratio": 1.5527638190954773,
            "no_speech_prob": 0.014556884765625,
            "words": [
                {
                    "start": 2945.01,
                    "end": 2945.73,
                    "word": " handicrafted",
                    "probability": 0.5687459309895834
                },
                {
                    "start": 2945.73,
                    "end": 2946.01,
                    "word": " social",
                    "probability": 0.75146484375
                },
                {
                    "start": 2946.01,
                    "end": 2946.87,
                    "word": " engineering",
                    "probability": 0.83984375
                },
                {
                    "start": 2946.87,
                    "end": 2947.31,
                    "word": " email",
                    "probability": 0.88720703125
                },
                {
                    "start": 2947.31,
                    "end": 2947.85,
                    "word": " attacks",
                    "probability": 0.60986328125
                },
                {
                    "start": 2947.85,
                    "end": 2948.55,
                    "word": ".",
                    "probability": 0.9052734375
                },
                {
                    "start": 2948.55,
                    "end": 2949.19,
                    "word": " In",
                    "probability": 0.8876953125
                },
                {
                    "start": 2949.19,
                    "end": 2949.37,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 2949.37,
                    "end": 2949.63,
                    "word": " future",
                    "probability": 0.94140625
                },
                {
                    "start": 2949.63,
                    "end": 2949.75,
                    "word": ",",
                    "probability": 0.7861328125
                },
                {
                    "start": 2949.75,
                    "end": 2949.87,
                    "word": " we",
                    "probability": 0.90869140625
                },
                {
                    "start": 2949.87,
                    "end": 2950.13,
                    "word": " want",
                    "probability": 0.467041015625
                },
                {
                    "start": 2950.13,
                    "end": 2950.27,
                    "word": " to",
                    "probability": 0.9140625
                },
                {
                    "start": 2950.27,
                    "end": 2950.63,
                    "word": " apply",
                    "probability": 0.82080078125
                },
                {
                    "start": 2950.63,
                    "end": 2950.85,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 2950.85,
                    "end": 2951.29,
                    "word": " same",
                    "probability": 0.81689453125
                }
            ]
        },
        {
            "id": 401,
            "seek": 286877,
            "start": 2951.29,
            "end": 2961.85,
            "text": " design decisions to new GPT-3 model. Thank you. Do you have any questions?",
            "tokens": [
                50744,
                1715,
                5327,
                281,
                777,
                26039,
                51,
                12,
                18,
                2316,
                13,
                1044,
                291,
                13,
                1144,
                291,
                362,
                604,
                1651,
                30,
                51160
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23833333094914755,
            "compression_ratio": 1.5527638190954773,
            "no_speech_prob": 0.014556884765625,
            "words": [
                {
                    "start": 2951.29,
                    "end": 2952.13,
                    "word": " design",
                    "probability": 0.93603515625
                },
                {
                    "start": 2952.13,
                    "end": 2952.65,
                    "word": " decisions",
                    "probability": 0.80224609375
                },
                {
                    "start": 2952.65,
                    "end": 2953.25,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 2953.25,
                    "end": 2953.85,
                    "word": " new",
                    "probability": 0.53759765625
                },
                {
                    "start": 2953.85,
                    "end": 2954.53,
                    "word": " GPT",
                    "probability": 0.9326171875
                },
                {
                    "start": 2954.53,
                    "end": 2954.89,
                    "word": "-3",
                    "probability": 0.99072265625
                },
                {
                    "start": 2954.89,
                    "end": 2955.73,
                    "word": " model",
                    "probability": 0.3798828125
                },
                {
                    "start": 2955.73,
                    "end": 2956.33,
                    "word": ".",
                    "probability": 0.8896484375
                },
                {
                    "start": 2958.77,
                    "end": 2959.49,
                    "word": " Thank",
                    "probability": 0.86962890625
                },
                {
                    "start": 2959.49,
                    "end": 2959.73,
                    "word": " you",
                    "probability": 0.8857421875
                },
                {
                    "start": 2959.73,
                    "end": 2960.19,
                    "word": ".",
                    "probability": 0.7275390625
                },
                {
                    "start": 2960.19,
                    "end": 2960.55,
                    "word": " Do",
                    "probability": 0.8984375
                },
                {
                    "start": 2960.55,
                    "end": 2960.61,
                    "word": " you",
                    "probability": 0.87060546875
                },
                {
                    "start": 2960.61,
                    "end": 2960.77,
                    "word": " have",
                    "probability": 0.88525390625
                },
                {
                    "start": 2960.77,
                    "end": 2960.97,
                    "word": " any",
                    "probability": 0.8203125
                },
                {
                    "start": 2960.97,
                    "end": 2961.49,
                    "word": " questions",
                    "probability": 0.89599609375
                },
                {
                    "start": 2961.49,
                    "end": 2961.85,
                    "word": "?",
                    "probability": 0.98681640625
                }
            ]
        },
        {
            "id": 402,
            "seek": 286877,
            "start": 2965.18,
            "end": 2969.6,
            "text": " If you do have any questions, please head over to the Discord channel.",
            "tokens": [
                51252,
                759,
                291,
                360,
                362,
                604,
                1651,
                11,
                1767,
                1378,
                670,
                281,
                264,
                32623,
                2269,
                13,
                51412
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23833333094914755,
            "compression_ratio": 1.5527638190954773,
            "no_speech_prob": 0.014556884765625,
            "words": [
                {
                    "start": 2965.18,
                    "end": 2965.9,
                    "word": " If",
                    "probability": 0.810546875
                },
                {
                    "start": 2965.9,
                    "end": 2965.98,
                    "word": " you",
                    "probability": 0.89697265625
                },
                {
                    "start": 2965.98,
                    "end": 2966.08,
                    "word": " do",
                    "probability": 0.87060546875
                },
                {
                    "start": 2966.08,
                    "end": 2966.24,
                    "word": " have",
                    "probability": 0.87158203125
                },
                {
                    "start": 2966.24,
                    "end": 2966.4,
                    "word": " any",
                    "probability": 0.79541015625
                },
                {
                    "start": 2966.4,
                    "end": 2966.84,
                    "word": " questions",
                    "probability": 0.90185546875
                },
                {
                    "start": 2966.84,
                    "end": 2967.02,
                    "word": ",",
                    "probability": 0.83837890625
                },
                {
                    "start": 2967.02,
                    "end": 2967.22,
                    "word": " please",
                    "probability": 0.728515625
                },
                {
                    "start": 2967.22,
                    "end": 2967.38,
                    "word": " head",
                    "probability": 0.865234375
                },
                {
                    "start": 2967.38,
                    "end": 2967.58,
                    "word": " over",
                    "probability": 0.86376953125
                },
                {
                    "start": 2967.58,
                    "end": 2967.78,
                    "word": " to",
                    "probability": 0.9052734375
                },
                {
                    "start": 2967.78,
                    "end": 2968.08,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 2968.08,
                    "end": 2968.4,
                    "word": " Discord",
                    "probability": 0.6953125
                },
                {
                    "start": 2968.4,
                    "end": 2968.88,
                    "word": " channel",
                    "probability": 0.75146484375
                },
                {
                    "start": 2968.88,
                    "end": 2969.6,
                    "word": ".",
                    "probability": 0.8779296875
                }
            ]
        },
        {
            "id": 403,
            "seek": 286877,
            "start": 2970.64,
            "end": 2974.98,
            "text": " They are currently there answering questions in the DEF CON Discord channel",
            "tokens": [
                51528,
                814,
                366,
                4362,
                456,
                13430,
                1651,
                294,
                264,
                10113,
                37,
                16596,
                32623,
                2269,
                51716
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23833333094914755,
            "compression_ratio": 1.5527638190954773,
            "no_speech_prob": 0.014556884765625,
            "words": [
                {
                    "start": 2970.64,
                    "end": 2971.36,
                    "word": " They",
                    "probability": 0.71875
                },
                {
                    "start": 2971.36,
                    "end": 2971.58,
                    "word": " are",
                    "probability": 0.8857421875
                },
                {
                    "start": 2971.58,
                    "end": 2971.94,
                    "word": " currently",
                    "probability": 0.72705078125
                },
                {
                    "start": 2971.94,
                    "end": 2972.46,
                    "word": " there",
                    "probability": 0.826171875
                },
                {
                    "start": 2972.46,
                    "end": 2972.82,
                    "word": " answering",
                    "probability": 0.57958984375
                },
                {
                    "start": 2972.82,
                    "end": 2973.4,
                    "word": " questions",
                    "probability": 0.90673828125
                },
                {
                    "start": 2973.4,
                    "end": 2973.64,
                    "word": " in",
                    "probability": 0.86279296875
                },
                {
                    "start": 2973.64,
                    "end": 2973.8,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 2973.8,
                    "end": 2974.04,
                    "word": " DEF",
                    "probability": 0.697998046875
                },
                {
                    "start": 2974.04,
                    "end": 2974.18,
                    "word": " CON",
                    "probability": 0.8623046875
                },
                {
                    "start": 2974.18,
                    "end": 2974.46,
                    "word": " Discord",
                    "probability": 0.716796875
                },
                {
                    "start": 2974.46,
                    "end": 2974.98,
                    "word": " channel",
                    "probability": 0.7451171875
                }
            ]
        },
        {
            "id": 404,
            "seek": 287480,
            "start": 2974.98,
            "end": 2980.64,
            "text": " under aiv-talks-text.",
            "tokens": [
                50392,
                833,
                257,
                592,
                12,
                29302,
                82,
                12,
                25111,
                13,
                50652
            ],
            "temperature": 0.0,
            "avg_logprob": -0.5621744791666666,
            "compression_ratio": 0.7241379310344828,
            "no_speech_prob": 0.1004638671875,
            "words": [
                {
                    "start": 2974.98,
                    "end": 2975.86,
                    "word": " under",
                    "probability": 0.301025390625
                },
                {
                    "start": 2975.86,
                    "end": 2976.3,
                    "word": " aiv",
                    "probability": 0.14019775390625
                },
                {
                    "start": 2976.3,
                    "end": 2979.16,
                    "word": "-talks",
                    "probability": 0.7913411458333334
                },
                {
                    "start": 2979.16,
                    "end": 2980.54,
                    "word": "-text",
                    "probability": 0.873779296875
                },
                {
                    "start": 2980.54,
                    "end": 2980.64,
                    "word": ".",
                    "probability": 0.9296875
                }
            ]
        }
    ],
    "text": " Hi, and today with Young-Hoo Lee, I'll be presenting Detecting Handcrafted Social Engineering Emails with a Bleeding Edge Neural Language Model. A bit about us before getting into the talk. Sophos is a security products and services company. We have a whole range of security products ranging from firewalls, to mobile security, to email security, to endpoint security. And the team that I manage, Sophos AI, is the AI shop inside of Sophos. We research and develop the company's machine learning technology. And then we're also responsible... Okay, I'm gonna mute myself. I'm gonna like definitely mute myself in here so I can listen to the talk. I'm happy if you need something. Young-Hoo Lee is one of the star researchers on my team. He's almost completely solely responsible for the machine learning subsystem that protects our Android customers, at least responsible for the R&D part of that work. He's also the principal author of the work presented today. I'm presenting in a supporting role here. I'll mostly just be setting the stage for the original work that Young-Hoo is gonna describe. So, the problem we're solving. Most generally, we're working on the problem of detecting phishing emails. More specifically, we're working on detecting business email compromise phishing emails and targeted phishing emails. So, to understand what this is, it's useful to say what it's not. We're not focused on detecting mass campaigns where detection reduces to a near-duplicate detection problem because attackers are just sending out millions of copies of basically the same phishing email. Detecting those types of emails turns out to be a side effect of our focus. But our real focus is on detecting new, custom-authored, bespoke phishing emails that are based on research on a target. And this diagram does a good job of getting across the kind of workflow we're looking at stopping. In this workflow, cyber criminals and attackers identify targets through open-source research, usually on the web. Then they establish contact with those targets. In step two, we call this grooming. In this case, they put out a lure, an initial email, usually or sometimes an initial text message, and then build trust and authenticity around some identity that they're impersonating with the mark. In step three, they cash out the trust that they've built with their targets and make an ask of those targets. Oftentimes, that will be around wiring money or sending credentials. And then in step four, they actually receive money or receive the credentials. So within targeted phishing, business email compromise, which is focused on stealing money from businesses and other organizations, has been a growing trend. So as you can see here, in July 2016, a few billion dollars were stolen, according to the FBI, through business email compromise attacks. These are targeted phishing attacks that extort money from organizations. In May 2017, that number had grown to almost six billion. And in 2018, that number had grown to more than 12 billion. I don't have data for the last two years, but I see no reason to believe that this trend has attenuated. I think it's likely it's continued on a similar trajectory. This is a big problem. We see this in Sophos's customer base, which is which is a pretty large sample. We also hear about it from other folks in the cybersecurity space. And so we're very focused on it because it's affecting people. And it's not just affecting large organizations. You can see on this on the axis on the right, that something like 80,000 organizations had been hit by July 2018 by these attacks. So there are lots of small and mid-sized organizations getting hit. And oftentimes, financial damage can be in the hundreds of thousands of dollars and really impact people's lives when these attacks happened. So again, just to reiterate, we're focused on primarily these business email compromise use cases, but also more generally targeted phishing in which a lot of manual labor goes into the phishing process on the criminal actors. And I think it almost goes without saying, but we're focused on step two and three of these criminal actors workflow, the steps that are mediated over email. So we see lots of malicious emails exchanged in the grooming step. And then we also see, obviously, a malicious email transmitted in the exchange of information step where the attacker makes the ask of the target, usually employee. And this is just because we're focused on email as our signal. And that's the scope of the work that I'll be talking about today. Just to flesh this out a little bit, here's an example, phishing email sent in the later epics of the grooming stage of an attacker's workflow. Here, the attacker has established themselves as an impersonated chancellor of UC Berkeley. They are emailing an employee at UC Berkeley, asking if they're available, looking to exchange messages with them, probably about to make an ask around a money transfer. Now, I think it's important to highlight why phishing detection is hard. And why in particular, detecting new previously unseen phishing emails written as part of a manual phishing campaign, like what I've just described, is hard. And what this boils down to, I think, is that classical natural language processing problems are hard. It's hard to get computers to understand language in any meaningful way and reason about language in any meaningful way. And detecting phishing emails really boils down to building models that have some level of understanding of language. So to understand why algorithmically it's hard to make sense of language, let's look at a few classical natural language processing problems. So one of those is coreference resolution. To get a sense of what this problem means, consider the following sentence. I went to the store for some milk, and based on the price, decided to buy it. Now, from a grammatical perspective, it could refer to the store or to the milk, right? It's possible that I went to the store for some milk, and based on the price of the store, I decided to buy it. But it's more likely that I bought the milk. As humans, in solving this coreference resolution problem, so resolving what it refers to, we have to plumb the depths of a number of complex mental models, right? We need to deploy our syntactical model of the English language, our semantic model of the English language, and our model of the world in which a person is more likely to have bought milk at the store than have bought the store itself. And that's how we solve this problem. Hopefully, it's clear that it's hard to get algorithms to do that. But to detect phishing emails, we really need to understand language. And this is a problem that's constituent of the problem of understanding language. Word polysemy is also a classic problem in algorithmic understanding of language. So a sentence like, he drank a lot and was quite the rake. Grammatically, it's valid to interpret this sentence as meaning that he drank a lot and was quite the garden tool used to rake leaves off your lawn. That's clearly not the right, that's not the sense in which the word rake is being used. The word rake is being used in the sense of a drunk, semi-criminal, sort of dissolute individual here. But it takes a pretty deep exercise of a human being's mental models to arrive at the sense in which this word was used. And not easy to reproduce in the form of an automated agent, either machine learning or based on regexes and rules. Sentiment detection is another hard problem in natural language processing. So consider the sentence, I'm not angry at all. No, of course. Why would I be angry that you spent our life savings on your mistress? So clearly, the speaker is angry here. And they're being sarcastic. Detecting that they're angry and sarcastic is not trivial and requires that we understand not only the syntactic structure of the sentence, but also the semantics of the sentence. And requires that we have the reflex that this person is probably angry if their interlocutor spent their life savings on their mistress. Okay, so to solve phishing means that we need algorithms that can make sense of language. Making sense of language is hard, as these three problems demonstrate. So a good solution to the phishing problem would model as intermediate steps to detecting that an email is a phishing email. At some level, be able to solve these problems and sit somewhere in the depths of this intermediate representation of the language that it's looking at. The other challenge we have, obviously, in any cybersecurity problem, at least any detection context in cybersecurity, is that we have adversaries who'd like to bypass our detection. And that's also worth considering. So these are all reasons why the problem that we're presenting here is a hard one, deserves... we haven't solved it completely, deserves attention from our community. Now, the approach that we're using to attack the phishing problem is based in neural networks and deep learning, and a specific advance that happened in the last few years known as transformers. So transformers, or more specifically, transformer blocks, are a new kind of construct in neural networks, much like convolutions were new, I think, as of the 90s or late 80s, and backpropagation was a new idea in neural networks, I think, starting in the 80s. Transformers are a new idea that's come out in recent years, and they help model language with a depth and fidelity that seems to be genuinely new and represent a step function in our ability to model language. So they're very exciting. The big idea behind the work that we're presenting today is that we're taking transformers and applying them to a cybersecurity problem, which we haven't seen much of before. So I'm going to talk a bit about what transformers are. A detailed discussion of how they work is beyond the scope of this talk, but I'm going to give some intuition. And then I'll pass the mic metaphorically over to Young, who will present on how we're using transformers. Okay, so here's an example, which I think helps to illustrate some ways in which transformers represent a real breakthrough in machine comprehension of language. So in this example, and I expect some of you have seen this because this went viral last year, a researcher wrote a prompt. This prompt is given at the top here in a shocking finding, that paragraph. And then a transformer model just took it from there and wrote a story based on that prompt. And I think when you see the story that the machine wrote, you'll see that it betrays a understanding of the syntax and semantics of language that's quite striking. So I think it's worth reading this out loud. The human-written prompt reads, in a shocking finding, scientists discovered a herd of unicorns living in a remote, previously unexplored valley in the Andes Mountains. Even more surprising to the researchers was the fact that the unicorn spoke perfect English. And now again, the challenge to the left off and write something coherent. And if it can, that reflects something around its ability to solve problems like the fundamental problems I described earlier. So the neural network continues. The scientists named the population after their distinctive horn, Ovid's unicorn. These four-horned, silver-white unicorns were previously unknown to science. Now, after almost two centuries, the mystery of what sparked this odd phenomenon is finally solved. Dr. Jorge Perez, an evolutionary biologist from the University of La Paz and several companions, were exploring the Andes Mountains when they found a small valley with no other animals or humans. Perez noticed that the valley had what appeared to be a natural fountain, surrounded by two peaks of rock and silver snow. Perez and the others then ventured further into the valley. By the time we reached the top of one peak, the water looked blue with some crystals on top, said Perez. And the essay continues. But what interested Younghoo and I in seeing examples of what transformers can do like this is that it seems that the transformers have, examples like these suggest that transformers have understood the syntax and semantics of the English language to some degree, and which begs the question, could the representations that the transformers learn in their parameter structure be useful in detecting phishing emails, since these transformer models seem to have distilled knowledge of how language works and what it means within their parameter settings. Okay, so this is one example of, sort of impressive example, of what transformers can do, that beg the question of whether or not they could be applied usefully to phishing. And we'll show results from our experiments later in Younghoo's section. Here's another example of just how sort of marvelous transformer models are with respect to how much more sophisticated they seem to be. So here a researcher typed into this box a description of some HTML that he would like the transformer models to generate. He writes, a button with the color of Donald Trump's hair. And the model actually writes, as a text completion, some HTML and CSS. It seems to understand that Donald Trump has yellow hair, and it makes a button, and it writes valid HTML and CSS. Which is, I think, just impressive in the absolute sense, but also for folks who've been in the NLP space for a while, it's impressive relatively. Five, ten years ago, if you would have shown somebody that we could do this in 2020, I think they would have been a bit incredulous. This is a big step for you. Applications like these represent a big step forward in natural language processing. And they're due, again, to this new idea called a transformer block. That's the key building block out of which models like this model, which is called GPT-3. It's from OpenAI, which is an AI research lab, are based. Okay. So, moving on. So, I want to go back a little bit in the history of NLP as a way of describing how transformers are new and different. So, I assume a substantial chunk of our audience today has studied basic machine learning on text. And typically, in like a machine learning 101 course, you learn about the bag of words model and about discrete states, Markov models of language. And I want to talk about these representations as a way of talking about how limited those representations were, and then I'll talk about how transformers break us free of some of those limitations. So, a bag of words model of a document is a way of representing a document numerically for the purposes of machine learning, in which we just count up how many times each word in that document appeared. And then we create a matrix out of all the documents we have, and all the words in all those documents. And the entries in those matrix are just word counts. So, in the case on the left over here, the column vectors in our matrix are documents, and they get one dimension per word in our vocabulary, and they get counts of how many times words appeared in that vocabulary in the entries of that particular vector. And hopefully it's intuitive that, you know, once you've represented your documents in a vocabulary space in this way, you can compare documents by taking some distance measure between pairs of documents. You can also train machine learning models on your document corpus to say classified news articles is about sports or politics. But what you've done in the first step of these models is drop out sequence information. So, you've forgotten about which words come in which order in the document, and you've just represented your document as a bag of words. Which is a useful simplifying assumption, and it's one that we still use today in some of the modeling we do in my research group at Sophos. But it throws out a ton of information that transformers and more modern models don't throw out. In the model on the right, we have a discrete state Markov model of language. Here each word is a state, and the concept of language given by the model is kind of a cheesier adventure story in which the next words in an utterance depends only on the current words, and you're just sort of drawing from a probability distribution and moving through this graph to generate language. You could never have generated anything close to that unicorn story using a Markov model, and yet as recently as the last five, ten years, there's lots of papers coming out around like using hidden Markov models to parse sentences and that kind of thing. These are still useful models, but we've gone far beyond the simplifying assumptions in these original sort of simplified models of the world in NLP from the past few decades. So let's contrast transformers now, and I'll get into more details about how transformers work in a second, but let's contrast transformers with these earlier natural language processing approaches. So pre-transformer, most machine learning approaches didn't consider words in context. Many approaches made the simplifying bag of words assumptions as kind of a first step in the modeling process, and then ran term vectors through models like topic models, or logistic regression, or support vector machines. Most models didn't model kind of like co-reference relationships between words, or sort of which words pertain to which other words in a sentence, and I'll talk about what that means later, but transformers do sort of solve that problem. Most approaches didn't operate on either words or characters. Typically the way people use transformers, we use well-chosen chunks of words, which allows us to model misspelling and this kind of thing, so there's been an improvement there in the current generation of natural language processing models. And older approaches tended not to use neural network technology. That's changed a lot in the best ideas that have come out of the neural network revolution that's been ongoing since around 2012. So transformers kind of kick apart a number of log jams in natural language modeling. They give words contextual representations, they model attention, like the relationship between words in a document, they use these smart partial word representations that allow for misspellings and just the tower of babble of vernaculars that appear like under the banner of, say, the English language on the internet, and they take advantage of ideas like residual connections and modern optimizers and many of the really good ideas that have come out of the neural network revolution. So these are all reasons why we wanted to test their applicability to phishing detection. Okay, so if you want to get into detail about how transformers work, I'd recommend this blog post by Jay Alomar. It's where I cribbed this screenshot here. I'm not going to get into the details of the series of matrix multiplications that comprise summations and various linear algebra and operations that comprise the transformer block. I just want to give a little bit of intuition here before passing the mic over to Younghoo. The basic idea behind a transformer block, which is kind of a Lego block, out of which you build a transformer-based neural network, is we're taking in a sequence of words. This diagram shows a very simple example where we're just taking a sequence of two words. Typically we've taken a larger window, like 512 words. We pass them into the block, and the way they get passed in is not just as entries in a term vector matrix, but actually as vectors themselves. The words get a vector representation. These are known as embeddings. We pass them both into the transformer network. The first thing that the transformer network does to our input word sequence is model the attention relationships between the words. In a two-word case, it's a little bit harder to describe here, but basically if you had a sequence of 15 words, for every word, the attention mechanism would compute how much that word pertains to the other 14 words in the sentence. You'll see how that has a relationship with the coreference resolution problem I talked about earlier. But it's just intuitive that there's a graph of word relationships in a sentence, and self-attention kind of models that in terms of which subjects pertain to which objects, which pronouns pertain to which people, etc. There's an addition and normalization step that happens when we've run this self-attention process a number of times and typically we don't just do self-attention once. We have a number of what are called heads and we run attention a number of times. We combine all that together, we do some non-linear transformation on it, and then we wind up with a new embedding of your original sequence of the same dimensionality as this original embedding, except that now thinking is encoded in this new representation in the context in which it appears, in the context of machines. Machines is encoded in the context of thinking. Then typically we stack these transformer blocks that we actually do. We have another transformer block that then sort of refines the representation and we keep going. Young will show that we use a number of transformer blocks in our phishing detection work in a few minutes. Here's some intuition as to what comes out of the attention mechanism in a typical transformer. So here's what a transformer block has decided it sort of relates to in an input sentence. So here we have the input sentence, the animal didn't cross the street because it was too tired. And the strongest attentional relationship here goes to the animal, which is interesting because one could ask whether or not it refers to the street or the animal here. One can interpret the weight of the connection to the animal, meaning that the transformer block has decided, in their quotes, that it pertains to the animal, which is really interesting. So hopefully you get some intuition as to how powerful this attentional representation is and how important it is in machine learning models, getting some what we might call understanding over the language that they're analyzing. Okay, so I want to put the intuition together around how transformers pertain to the work that YoungHu and I are presenting today. So basically, what we're going to do in our phishing model is embed an email as a sequence of embedded character sequence vectors. And then we're going to run that through a series of transformer blocks, like what I just showed, that are going to create a very refined attentional representation of the word sequences, and then produce these contextual embeddings that get at the meaning of the words in the context in which they appear. And then finally, our network is going to solve a classification task, say whether or not the email is a phishing email or not. How all that magic works, how the network gets trained, which there's some tricks there that are really cool. I'll leave to YoungHu, and hope a lot of this makes sense, and happy to take questions about my piece of this presentation later at the end of the talk. Thank you, Joshi. Let me continue the second part of our talk. The second part will include our design decisions for CatBot and performance result. CatBot is the name of our email model, context-aware tiny bot. The model size is tiny, but it is mighty bot. Modern NLP models all have a nice and friendly name. For example, Elmer was introduced in 2018. The model used bidirectional LSTMs to generate contextualized word embeddings. And then, same year later, Google researchers introduced bot and achieved the state-of-art performance in many English-understanding problems. Next year, 2019, Baidu researchers introduced the Orni, and the model achieved another state-of-art performance in many Chinese-language-understanding problems. They are all popular characters from Sesame Street. This year, 2020, we introduced CatBot to tackle email security problems. Transformer-based NLP models are powerful, but they are complex and heavy. And it is challenging to deploy heavy models for real-time applications. So our first design goal is to convert the heavy model into a lightweight model, so we can reduce the number of parameters and then we can improve inference speed. We downsized a baseline model called DistribBot, which has six transformer blocks. We take half of transformer blocks from pre-trained model and then replace missing transformers with simple adapters. For example here, we take transformer 1, 3, 5, and then we added two adapters. Also, we can take other number of transformer blocks. This approach allows simply significantly reduced number of parameters. The second goal is to improve the model performance by combining additional input. Standard NLP models only accept text data as input. However, we can extract additional features from email headers and we can use the additional input to our model. So the text input will be the input to the embedding and then we can add additional input to the classification header and we added additional dense layers in the classification to combine the input from transformer block and another header-related input. With the additional input, we improved our model's performance further. Let me talk about the details of our trans adapters. We inserted two adapters here and each adapter has quite a simple architecture. Each adapter will have two dense layers and there is one non-linear activation unit in between and we have a skip connection. The dimensionality of a dense unit is the same as the output of transformer blocks. And the two dense layers are initialized with near-zero values. So as initially, the adapters will act as identity blocks, but however, they will gradually change the data from the lower transformer block to upper transformer block and to minimize classification loss. We also modified the standard fine-tuning method by using a partial fine-tuning method. Standard fine-tuning involves updating all parameters jointly. However, for partial fine-tuning, we only update upper blocks, but we fix low blocks. For example, here, the low blocks embedding the transformer 1 and 3 are fixed, but we update adapter 1 and 2 and transformer 5 and classification header. This approach was to minimize forgetting problems of learned presentations from low transformer blocks. As mentioned earlier, we have two set of features. One from text data and another one from email headers. We can use multiple email header builders, for example, from two cc-reply builders to extract additional context information. And we consider the subject and the text as context input. The first set of features are from email text content features. We extract text data from subject and plain text body. If only html content is available, then we can also extract plain text from html data using a html parser. For example, this one is a simple html hyperlink, but we only extract visible text visit bank site as output. And then from the extracted text, we remove less informative characters. We remove digit and punctuation characters. The second example, there are many hyphens, but the hyphens will be removed in this step. Otherwise, each hyphen will be individual token. Finally, we select 120 tokens as input to the transformer blocks. We use a sub-word tokenized called WordPiece. The WordPiece tokenized can overcome some of the limitations from character-level or word-level tokens. Character-level tokens are too fine-grained, so it is hard to recognize word boundaries and meaning of words. And word-level tokens often have auto-vocabulary problems. The sub-word token tokenized can split complex or uncommon words into sub-tokens. For example, CAPBOT can be divided into simple CAP and BOT, and double HES is an indicator for sub-words. Similarly, SOPHOS can be divided into three tokens, SO, PH, and OS. The sub-word tokenized reduce number of unknown tokens in our email data. With the selected tokens, the tokens will be the input to the embedding layer, and then we have three transformer blocks. And each transformer block has 12 multi-head attention layers. And each individual attention layer runs contextual relationship between tokens. And the lighter diagram shows attention rate between tokens. The transfer token has multiple attention rate for other tokens. In our email data, we have many non-English emails, and also non-English emails can include English words. Also, English emails can include non-English words. The non-English emails account for 25% of our total P9 and malicious emails. So, we needed to support a multilingual model which can recognize different languages. How we can support multilingual emails? The solution is BOT comes with two versions, English and multilingual BOT. The English version was pre-trained with the English text datasets, including Wikipedia, and has 30,000 English tokens. The multilingual version was pre-trained with large text datasets from more than 100 languages. And this version has four times large vocabulary size, which is 120 tokens, which will cover many Unicode characters and Unicode words. So, we fine-tuned a multilingual BOT for our multilingual emails. The second set of results from email headers. We can extract multiple indicators from email header builders. For example, the first one, we check whether the emails are from internal or external. We can compare the domain of recipient and sender. For example, here, the domain name is a similar looking one, but actually there are extra ads. So, this one we consider as external email. And then, we can also use external reply by comparing the domain of from and reply to. Often, targeted phishing attackers use a different domain for reply to. And also, we collect the size of the recipients and the size of the carbon copy recipients as additional indicators. It is obvious that targeted phishing attacks will have only single recipient. However, many user phishing emails will have multiple recipients or carbon copy recipients. Next, we will have a look at the performance of our CatBot. Let's have a look at the performance of a CatBot. We used a data set of 10 million benign samples. And the data set also includes 350 phishing emails and 1000 BC emails. We used time split to allocate 70% of samples as training and remaining 30% for test samples. We included two baseline models to compare CatBot. The first one is DistilBot, which has six transformer blocks. And another one is LSTM Long Short Memory. The model is a recurrent neural network architecture, which also uses BERT's same embedding layer. We trained three models on a GPU instance from AWS. And we assigned high sample weight for the PEC samples and oversample minor class malicious samples to allocate balanced samples in each mini-matching. To compare performance, we use R-Curves and L-Re-Under-Curve. Also, we compare inference speed and the model size as key performance metrics. These R-Curves compare our CatBot model with two baseline models. The top blue one is CatBot, and the second one is DistilBot, and the bottom green one is LSTM. Our model outperformed the two baseline models. And our model achieved 0.82% positive rate at 0.1% false positive rate. Next, we compare the performance when we removed adapters and context input rated layers. The top one is CatBot, and the second orange one is when we removed adapters. And the bottom one is when context input was removed from the CatBot. We can see significant performance drop when we remove either adapter or context input, which demonstrates we can improve performance by using additional adapters and context rated layers. Next, we compare the performance of three models with targeted BC samples. We assigned high sample rate for BC samples, and we achieved high performance for detecting those BC samples and the CatBot above the two baseline models. Next, we compare performance for phishing emails. We divide the phishing emails into two groups, English and non-English emails. Our CatBot outperformed for English and non-English emails. And our model was based on the multilingual bot, so we can see significant performance when we use the model for detecting non-English emails, when it compares with simple LST model. Next, we compare the inference speed. This bot has six transformer blocks, and the CatBot has three transformer blocks, so we achieve two times speed up in inference time when we measure the performance on a CPU machine. As the number of blocks decrease, the inference time can be reduced. Next, model size. For comparison, we divide the model size into two parts, embedding and transformer blocks. The DistilBot has six transformers and has 92 million parameters for embedding and 42 million parameters for transformer blocks. DistilBot and the CatBot have six transformer blocks. We reuse the same embedding parameters, but we reduce the number of parameters for transformer blocks by 50%. In total, our model size is 85% of baseline model. When we apply the same mechanism for English version, the English CatBot will have 71% of parameters from baseline model. Next, we will inspect how CatBot generates outputs. We use a LIME method to interpret our predictions. LIME is a local interpretable model ergodastic explanation method which can be applied to any black box models. We can understand a model by perturbing the input and understanding how the predictions change. The first LIME example is for a benign email. The prediction score for this email is close to zero. And we highlight legitimate tokens with blue color and malicious ones as orange one. And this one, we don't have any high-rated tokens for malicious ones. Next, we have a BC sample. The model prediction score for maliciousness is close to one. And the model recognizes transfer and urgent payment as high-rated tokens. Next, we have another BC sample, which is related with gift card. And the model prediction score is close to one and card and urgently. The tokens are high-rated for this email. Next, we have two handcrafted social engineering emails. They look quite different, but if you read the text carefully, they are asking the same wire transfer. And the model predictions are close to one for both emails and the highlight tokens are payment and swift or as soon as possible. These examples demonstrated our model's ability to understand complex texts and conceptually similar emails can be identified. In conclusion, our CatBot is a carefully re-architectural transform-based model. With this architecture, we achieve both high-speed and high-accuracy in detecting handicrafted social engineering email attacks. In the future, we want to apply the same design decisions to new GPT-3 model. Thank you. Do you have any questions? If you do have any questions, please head over to the Discord channel. They are currently there answering questions in the DEF CON Discord channel under aiv-talks-text."
}