audit_report.json

{
  "pipeline_stages": [
    [
      "Stage 1",
      "Baseline",
      517246,
      0
    ],
    [
      "Stage 2",
      "Encoding",
      478366,
      38880
    ],
    [
      "Stage 3",
      "Structural Noise",
      467628,
      10738
    ],
    [
      "Stage 4",
      "Boilerplate",
      451427,
      16201
    ],
    [
      "Stage 5",
      "Segmentation",
      457907,
      3792
    ],
    [
      "Stage 6",
      "Exact Dedup",
      440917,
      16990
    ],
    [
      "Stage 7",
      "Near Dedup",
      435595,
      5322
    ],
    [
      "Stage 8",
      "Script Filter",
      413162,
      22433
    ],
    [
      "Stage 9",
      "Normalization",
      413162,
      0
    ],
    [
      "Stage 10",
      "Length Filter",
      389223,
      23939
    ],
    [
      "Stage 11",
      "Quality Selection",
      389223,
      0
    ]
  ],
  "final_statistics": {
    "sentence_count": 389223,
    "total_words": 6342454,
    "vocabulary_size": 498874,
    "type_token_ratio": 0.078656305587711,
    "median_chars": 92.0,
    "median_words": 15.0,
    "mean_sinhala_ratio": 0.9654833341843978
  }
}