{
  "text": "MinerU Popo是一個面向複雜文檔解析的後處理模型框架。核心目標是把OCR和版面分析得到的頁面集結果,進一步整理成文檔集結構化知識。\n它並非替代MinerU、Manty OCR或PaddleOCR等項目,而是在這些解析工具進行OCR之後,對標題層級、段落連續性、跨頁表格、圖文關係和閱讀順序進行統一推理與修復。\n通俗地說,OCR能力是負責把字看出來。MinerU Popo負責把整份文檔真正組織起來。\n在技術上,MinerU Popo採用4B參數後處理模型,並集合任務化數據引擎、動態分塊同步和文檔增強方法。\n重點解決長文檔在分頁、跨頁和複雜版式下容易出現的結構斷裂問題。\n在公開指標中,接入Popo框架後,MinerU的TEDS從53.7提升到90.6。\nMineQ-C2從48.9提升到87.4,說明它能夠顯著提升文檔結構恢復質量。\n它的價值在於讓知識庫、RAG智能問答、學術論文解析和企業文檔治理,獲得更乾淨、更穩定、更可計算的文檔資產。\nMinerU Popo讓複雜文檔從可識別,走向可理解、可治理、可計算。\nMinerU Popo",
  "language": "zh",
  "duration": 80.828688,
  "segments": [
    {
      "id": 0,
      "seek": 0,
      "start": 0.0,
      "end": 11.4,
      "text": "MinerU Popo是一個面向複雜文檔解析的後處理模型框架。核心目標是把OCR和版面分析得到的頁面集結果,進一步整理成文檔集結構化知識。",
      "tokens": [
        50365,
        44,
        4564,
        52,
        10215,
        78,
        1541,
        8990,
        8833,
        24282,
        164,
        97,
        229,
        6306,
        250,
        17174,
        40741,
        242,
        17278,
        7360,
        238,
        1546,
        5661,
        39289,
        13876,
        41908,
        39823,
        21416,
        228,
        7360,
        114,
        1543,
        8987,
        116,
        7945,
        11386,
        46299,
        1541,
        16075,
        46,
        18547,
        12565,
        42096,
        8833,
        6627,
        7360,
        238,
        5916,
        4511,
        1546,
        8313,
        223,
        8833,
        26020,
        35181,
        11,
        18214,
        2257,
        31429,
        27662,
        13876,
        11336,
        17174,
        40741,
        242,
        26020,
        17144,
        43362,
        23756,
        6498,
        43143,
        1543,
        50935
      ],
      "temperature": 0,
      "avg_logprob": -0.15639057514830407,
      "compression_ratio": 1.1832298136645962,
      "no_speech_prob": 9.6452212536291e-12,
      "source": "whisper"
    },
    {
      "id": 1,
      "seek": 0,
      "start": 11.4,
      "end": 25.8,
      "text": "它並非替代MinerU、Manty OCR或PaddleOCR等項目,而是在這些解析工具進行OCR之後,對標題層級、段落連續性、跨頁表格、圖文關係和閱讀順序進行統一推理與修復。",
      "tokens": [
        50935,
        11284,
        35934,
        12107,
        9531,
        123,
        19105,
        44,
        4564,
        52,
        1231,
        44,
        394,
        88,
        422,
        18547,
        19780,
        50112,
        2285,
        46,
        18547,
        10187,
        8313,
        227,
        11386,
        11,
        11070,
        1541,
        3581,
        29869,
        17278,
        7360,
        238,
        23323,
        39806,
        18214,
        8082,
        46,
        18547,
        26362,
        11,
        2855,
        46299,
        14475,
        9636,
        97,
        35315,
        11,
        28427,
        30848,
        36252,
        29583,
        21686,
        1231,
        6563,
        101,
        8313,
        223,
        17571,
        30921,
        1231,
        2523,
        244,
        17174,
        25802,
        12565,
        29602,
        7422,
        222,
        8313,
        228,
        6346,
        237,
        18214,
        8082,
        33725,
        2257,
        33597,
        13876,
        34127,
        7792,
        106,
        2172,
        102,
        1543,
        51655
      ],
      "temperature": 0,
      "avg_logprob": -0.15639057514830407,
      "compression_ratio": 1.1832298136645962,
      "no_speech_prob": 9.6452212536291e-12,
      "source": "whisper",
      "text_original": "它並非替代MinerU、Manty OCR或PaddleOCR等項目,而是在這些解析工具進行OCR之後,對標題層級,段落連續性、跨頁表格、圖文關係和閱讀順序進行統一推理與修復。",
      "correction_applied": true,
      "correction_source": "llm_correct_a_presegmentation"
    },
    {
      "id": 2,
      "seek": 2580,
      "start": 25.8,
      "end": 32.6,
      "text": "通俗地說,OCR能力是負責把字看出來。MinerU Popo負責把整份文檔真正組織起來。",
      "tokens": [
        50365,
        19550,
        7792,
        245,
        10928,
        4622,
        11,
        46,
        18547,
        8225,
        13486,
        1541,
        11561,
        254,
        11561,
        105,
        16075,
        22381,
        4200,
        29741,
        1543,
        44,
        4564,
        52,
        10215,
        78,
        11561,
        254,
        11561,
        105,
        16075,
        27662,
        36266,
        17174,
        40741,
        242,
        6303,
        15789,
        30052,
        23141,
        242,
        21670,
        1543,
        50705
      ],
      "temperature": 0,
      "avg_logprob": -0.10067456994353041,
      "compression_ratio": 1.191044776119403,
      "no_speech_prob": 1.7110691336030825e-11,
      "source": "whisper"
    },
    {
      "id": 3,
      "seek": 2580,
      "start": 32.6,
      "end": 42.400000000000006,
      "text": "在技術上,MinerU Popo採用4B參數後處理模型,並集合任務化數據引擎、動態分塊同步和文檔增強方法。",
      "tokens": [
        50705,
        3581,
        32502,
        9890,
        241,
        5708,
        11,
        44,
        4564,
        52,
        10215,
        78,
        6900,
        94,
        9254,
        19,
        33,
        47375,
        30622,
        5661,
        39289,
        13876,
        41908,
        39823,
        11,
        35934,
        26020,
        14245,
        26443,
        39517,
        23756,
        30622,
        36841,
        41301,
        13167,
        236,
        1543,
        12572,
        45248,
        6627,
        26268,
        13089,
        31429,
        12565,
        17174,
        40741,
        242,
        24228,
        252,
        22752,
        9249,
        11148,
        1543,
        51195
      ],
      "temperature": 0,
      "avg_logprob": -0.10067456994353041,
      "compression_ratio": 1.191044776119403,
      "no_speech_prob": 1.7110691336030825e-11,
      "source": "whisper",
      "text_original": "在技術上,MinerU Popo採用4B參數後處理模型,並集合任務化數據引擎。動態分塊同步和文檔增強方法。",
      "correction_applied": true,
      "correction_source": "llm_correct_a_presegmentation"
    },
    {
      "id": 4,
      "seek": 2580,
      "start": 42.400000000000006,
      "end": 48.400000000000006,
      "text": "重點解決長文檔在分頁、跨頁和複雜版式下容易出現的結構斷裂問題。",
      "tokens": [
        51195,
        12624,
        8216,
        17278,
        33540,
        15353,
        17174,
        40741,
        242,
        3581,
        6627,
        18573,
        120,
        1231,
        6563,
        101,
        18573,
        120,
        12565,
        164,
        97,
        229,
        6306,
        250,
        42096,
        27584,
        4438,
        49212,
        7781,
        9581,
        1546,
        17144,
        43362,
        4307,
        115,
        8083,
        224,
        17197,
        1543,
        51495
      ],
      "temperature": 0,
      "avg_logprob": -0.10067456994353041,
      "compression_ratio": 1.191044776119403,
      "no_speech_prob": 1.7110691336030825e-11,
      "source": "whisper",
      "text_original": "重點解決長文檔在分翼、跨翼和複雜版式下容易出現的結構斷裂問題。",
      "correction_applied": true,
      "correction_source": "llm_correct_a_presegmentation"
    },
    {
      "id": 5,
      "seek": 2580,
      "start": 48.400000000000006,
      "end": 55.400000000000006,
      "text": "在公開指標中,接入Popo框架後,MinerU的TEDS從53.7提升到90.6。",
      "tokens": [
        51495,
        3581,
        13545,
        8949,
        25922,
        46299,
        5975,
        11,
        14468,
        14028,
        38677,
        78,
        21416,
        228,
        7360,
        114,
        5661,
        11,
        44,
        4564,
        52,
        1546,
        51,
        4731,
        50,
        21068,
        19584,
        13,
        22,
        20949,
        41670,
        4511,
        7771,
        13,
        21,
        1543,
        51845
      ],
      "temperature": 0,
      "avg_logprob": -0.10067456994353041,
      "compression_ratio": 1.191044776119403,
      "no_speech_prob": 1.7110691336030825e-11,
      "source": "whisper"
    },
    {
      "id": 6,
      "seek": 5540,
      "start": 55.4,
      "end": 62.4,
      "text": "MineQ-C2從48.9提升到87.4,說明它能夠顯著提升文檔結構恢復質量。",
      "tokens": [
        50365,
        44,
        533,
        48,
        12,
        34,
        17,
        21068,
        13318,
        13,
        24,
        20949,
        41670,
        4511,
        23853,
        13,
        19,
        11,
        4622,
        11100,
        11284,
        8225,
        31649,
        46605,
        19382,
        20949,
        41670,
        17174,
        40741,
        242,
        17144,
        43362,
        11845,
        95,
        2172,
        102,
        43450,
        26748,
        1543,
        50715
      ],
      "temperature": 0,
      "avg_logprob": -0.12722345079694475,
      "compression_ratio": 1.171003717472119,
      "no_speech_prob": 1.9842270188030753e-11,
      "source": "whisper"
    },
    {
      "id": 7,
      "seek": 5540,
      "start": 62.4,
      "end": 72.4,
      "text": "它的價值在於讓知識庫、RAG智能問答、學術論文解析和企業文檔治理,獲得更乾淨、更穩定、更可計算的文檔資產。",
      "tokens": [
        50715,
        45224,
        45441,
        40242,
        3581,
        19488,
        21195,
        6498,
        43143,
        6346,
        104,
        1231,
        3750,
        38,
        5094,
        118,
        8225,
        11361,
        28223,
        1231,
        21372,
        9890,
        241,
        29340,
        17174,
        17278,
        7360,
        238,
        12565,
        7384,
        223,
        27119,
        17174,
        40741,
        242,
        30118,
        13876,
        11,
        29057,
        110,
        5916,
        19002,
        39698,
        14176,
        101,
        1231,
        19002,
        13740,
        102,
        12088,
        1231,
        19002,
        4429,
        30114,
        19497,
        1546,
        17174,
        40741,
        242,
        35380,
        33299,
        1543,
        51215
      ],
      "temperature": 0,
      "avg_logprob": -0.12722345079694475,
      "compression_ratio": 1.171003717472119,
      "no_speech_prob": 1.9842270188030753e-11,
      "source": "whisper"
    },
    {
      "id": 8,
      "seek": 5540,
      "start": 72.4,
      "end": 78.4,
      "text": "MinerU Popo讓複雜文檔從可識別,走向可理解、可治理、可計算。",
      "tokens": [
        51215,
        44,
        4564,
        52,
        10215,
        78,
        21195,
        164,
        97,
        229,
        6306,
        250,
        17174,
        40741,
        242,
        21068,
        4429,
        43143,
        16158,
        11,
        9575,
        24282,
        4429,
        13876,
        17278,
        1231,
        4429,
        30118,
        13876,
        1231,
        4429,
        30114,
        19497,
        1543,
        51515
      ],
      "temperature": 0,
      "avg_logprob": -0.12722345079694475,
      "compression_ratio": 1.171003717472119,
      "no_speech_prob": 1.9842270188030753e-11,
      "source": "whisper"
    },
    {
      "id": 9,
      "seek": 7840,
      "start": 78.4,
      "end": 80.4,
      "text": "MinerU Popo",
      "tokens": [
        50365,
        44,
        4564,
        52,
        10215,
        78,
        50465
      ],
      "temperature": 0,
      "avg_logprob": -0.17050259009651517,
      "compression_ratio": 12.434782608695652,
      "no_speech_prob": 1.3218524053115743e-09,
      "source": "whisper"
    }
  ],
  "transcript_selection": {
    "schema": "videonote.transcript_selection.v1",
    "status": "SELECTED",
    "requested_language": "zh",
    "duration": 80.828688,
    "reference_path": null,
    "reference_kind": null,
    "reference_language": null,
    "selected_source": "whisper_segmented",
    "decision_reason": "source_unavailable_or_untrusted_whisper_passed",
    "whisper_rejected": false,
    "source_profile": {
      "segment_count": 0,
      "duration_coverage": 0.0,
      "usable_full": false,
      "usable_partial": false,
      "hallucination_region_count": 0,
      "hallucination_regions": [],
      "language": {
        "requested": "zh",
        "matched": false
      }
    },
    "whisper_profile": {
      "segment_count": 10,
      "multilingual": false,
      "first_start": 0.0,
      "final_end": 80.4,
      "duration_coverage": 0.994696,
      "maximum_gap_seconds": 0.0,
      "inverted_count": 0,
      "language": {
        "requested": "zh",
        "matched": true,
        "latin": 106,
        "han": 299,
        "kana": 0,
        "hangul": 0
      },
      "hallucination_region_count": 0,
      "hallucination_regions": [],
      "usable_full": true,
      "usable_partial": true
    },
    "comparison": {
      "window_seconds": 30.0,
      "compared_window_count": 0,
      "median_score": 0.0,
      "divergent_window_count": 0,
      "divergent_ratio": 0.0,
      "divergent_windows": []
    },
    "timestamp_alignment": {
      "applied": false,
      "estimated_offset_seconds": 0.0,
      "matched_segment_count": 0,
      "match_scores": []
    },
    "caption_timing_repair": null,
    "filled_source_profile": null,
    "whisper_duration_guard": {
      "media_duration": 80.829,
      "rejected_segment_count": 25,
      "rejected_segments": [
        {
          "index": 10,
          "start": 80.4,
          "end": 81.4,
          "reason": "material_end_timestamp_overrun"
        },
        {
          "index": 11,
          "start": 81.4,
          "end": 82.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 12,
          "start": 82.4,
          "end": 83.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 13,
          "start": 83.4,
          "end": 84.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 14,
          "start": 84.4,
          "end": 85.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 15,
          "start": 85.4,
          "end": 86.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 16,
          "start": 86.4,
          "end": 87.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 17,
          "start": 87.4,
          "end": 88.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 18,
          "start": 88.4,
          "end": 89.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 19,
          "start": 89.4,
          "end": 91.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 20,
          "start": 91.4,
          "end": 92.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 21,
          "start": 92.4,
          "end": 93.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 22,
          "start": 93.4,
          "end": 94.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 23,
          "start": 94.4,
          "end": 95.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 24,
          "start": 95.4,
          "end": 96.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 25,
          "start": 96.4,
          "end": 97.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 26,
          "start": 97.4,
          "end": 98.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 27,
          "start": 98.4,
          "end": 99.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 28,
          "start": 99.4,
          "end": 100.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 29,
          "start": 100.4,
          "end": 101.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 30,
          "start": 101.4,
          "end": 102.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 31,
          "start": 102.4,
          "end": 103.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 32,
          "start": 103.4,
          "end": 104.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 33,
          "start": 104.4,
          "end": 105.4,
          "reason": "starts_at_or_after_media_end"
        },
        {
          "index": 34,
          "start": 105.4,
          "end": 106.4,
          "reason": "starts_at_or_after_media_end"
        }
      ]
    }
  },
  "presegmentation_correction": {
    "applied": true,
    "correction_count": 3,
    "deterministic_fix_count": 0
  }
}