audit_report.json

{
  "pipeline_stages": [
    [
      "Stage 1",
      "Baseline",
      7281026,
      0
    ],
    [
      "Stage 2",
      "Encoding",
      7105977,
      175049
    ],
    [
      "Stage 3",
      "Structural Noise",
      6997728,
      108249
    ],
    [
      "Stage 4",
      "Boilerplate",
      6509247,
      488481
    ],
    [
      "Stage 5",
      "Segmentation",
      6570046,
      19077
    ],
    [
      "Stage 6",
      "Exact Dedup",
      6374926,
      195120
    ],
    [
      "Stage 7",
      "Near Dedup",
      6283262,
      91664
    ],
    [
      "Stage 8",
      "Script Filter",
      5944633,
      338629
    ],
    [
      "Stage 9",
      "Normalization",
      5944633,
      0
    ],
    [
      "Stage 10",
      "Length Filter",
      5684456,
      260177
    ],
    [
      "Stage 11",
      "Quality Selection",
      1000000,
      4684456
    ]
  ],
  "final_statistics": {
    "sentence_count": 1000000,
    "total_words": 10746623,
    "vocabulary_size": 520446,
    "type_token_ratio": 0.048428794794420534,
    "median_chars": 64.0,
    "median_words": 11.0,
    "mean_sinhala_ratio": 0.9955517945507983
  },
  "quality_scoring": {
    "total_scored": 5684456,
    "target_size": 1000000,
    "cutoff_score": 0.9790849902188197,
    "score_min": 0.5027132658029089,
    "score_median": 0.9296862943770501,
    "score_max": 1.0
  }
}