{
  "by_reference_syllables": [
    {
      "cohort": 15106,
      "exact": 12999,
      "exact_percent": 86.05,
      "over_counted": 2107,
      "reference_syllables": 1,
      "under_counted": 0
    },
    {
      "cohort": 54064,
      "exact": 49192,
      "exact_percent": 90.99,
      "over_counted": 3108,
      "reference_syllables": 2,
      "under_counted": 1764
    },
    {
      "cohort": 32395,
      "exact": 26136,
      "exact_percent": 80.68,
      "over_counted": 1541,
      "reference_syllables": 3,
      "under_counted": 4718
    },
    {
      "cohort": 11739,
      "exact": 8191,
      "exact_percent": 69.78,
      "over_counted": 476,
      "reference_syllables": 4,
      "under_counted": 3072
    },
    {
      "cohort": 3342,
      "exact": 2135,
      "exact_percent": 63.88,
      "over_counted": 100,
      "reference_syllables": 5,
      "under_counted": 1107
    },
    {
      "cohort": 712,
      "exact": 398,
      "exact_percent": 55.9,
      "over_counted": 14,
      "reference_syllables": 6,
      "under_counted": 300
    },
    {
      "cohort": 110,
      "exact": 52,
      "exact_percent": 47.27,
      "over_counted": 1,
      "reference_syllables": 7,
      "under_counted": 57
    },
    {
      "cohort": 15,
      "exact": 5,
      "exact_percent": 33.33,
      "over_counted": 0,
      "reference_syllables": 8,
      "under_counted": 10
    },
    {
      "cohort": 2,
      "exact": 0,
      "exact_percent": 0.0,
      "over_counted": 0,
      "reference_syllables": 9,
      "under_counted": 2
    },
    {
      "cohort": 1,
      "exact": 0,
      "exact_percent": 0.0,
      "over_counted": 0,
      "reference_syllables": 12,
      "under_counted": 1
    }
  ],
  "cohort_definition": "Every all-ASCII letter spelling with a positive primary syllable count in the compiled pronunciation release; pronunciation is hidden from the estimator.",
  "data_version": "2026.07-lexical.6",
  "limitations": "The source-covered cohort is not representative of unseen names, coinages, dialects, or multilingual text. Variant pronunciations are not separate rows.",
  "method": "Run the public spelling fallback on each spelling, compare it with the retained primary pronunciation count, and report exact, within-one, under-, and over-counts.",
  "metrics": {
    "cohort": 117486,
    "exact": 99108,
    "exact_percent": 84.36,
    "over_counted": 7347,
    "over_counted_percent": 6.25,
    "under_counted": 11031,
    "under_counted_percent": 9.39,
    "within_one": 116884,
    "within_one_percent": 99.49
  },
  "schema_version": 1,
  "source_file": "pronunciations.jsonl",
  "source_sha256": "30fe7942dae60ab07b5cae443b06d47d3cb71b83e9d5181e2b4c2ec6b6522fce",
  "study_id": "syllable-spelling-fallback-masked-reference-v1",
  "title": "Syllable spelling-fallback masked-reference benchmark"
}