audit_report.json

{
  "pipeline_stages": [
    [
      "Stage 1",
      "Baseline",
      5415583,
      0
    ],
    [
      "Stage 2",
      "Encoding",
      5383263,
      32320
    ],
    [
      "Stage 3",
      "Structural Noise",
      5383263,
      55038
    ],
    [
      "Stage 4",
      "Boilerplate",
      5124469,
      203756
    ],
    [
      "Stage 5",
      "Segmentation",
      5237425,
      22673
    ],
    [
      "Stage 6",
      "Exact Dedup",
      4354184,
      883241
    ],
    [
      "Stage 7",
      "Near Dedup",
      4256075,
      98109
    ],
    [
      "Stage 8",
      "Script Filter",
      4200134,
      55941
    ],
    [
      "Stage 9",
      "Normalization",
      4200134,
      0
    ],
    [
      "Stage 10",
      "Length Filter",
      3547117,
      653017
    ],
    [
      "Stage 11",
      "Quality Selection",
      1000000,
      2547117
    ]
  ],
  "final_statistics": {
    "sentence_count": 1000000,
    "total_words": 16725477,
    "vocabulary_size": 455455,
    "type_token_ratio": 0.027231211402819783,
    "median_chars": 106.0,
    "median_words": 17.0,
    "mean_sinhala_ratio": 0.9946439881001835
  },
  "quality_scoring": {
    "total_scored": 3547117,
    "target_size": 1000000,
    "cutoff_score": 0.9745396313703123,
    "score_min": 0.5192261459110801,
    "score_median": 0.9480029154018438,
    "score_max": 1.0
  }
}