NIGAMI LAB データセット

ファイルはすべて UTF-8 の CSV（1 行目が列名）。id で相互に参照します。
- entities.csv: 項目（食材・生物・化合物・受容体・感覚・工程・料理・調味料・スパイス・地域・文化・概念）
- names.csv: 項目の名称（言語・文字・種別。script=Hira かつ notes が「読み仮名」で始まる行は並び替え用の読み）
- sources.csv: 出典台帳（review_status が確認範囲: full_text_checked / abstract_checked / metadata_only）
- claims.csv: 主張（subject → predicate → object。claim_status が確度: established / supported / preliminary / contested / contextual / observation / unverified）
- evidence.csv: 主張と出典の接続（relation: supports / contradicts / limits / contextualizes。finding_paraphrase は資料の要約、limitations は限定）
- entity-overviews.csv / entity-overview-evidence.csv: 項目の本文段落と、その段落が使った根拠
- measurements.csv: 閾値・濃度などの数値
- recipes.csv: 料理の構成（verification_status）
- leads.csv: 仮説採集箱（問いとして保持しているもの）
- source-checks.csv: 出典の確認の記録（いつ・どこまで読んだか。scope=全文/部分/抄録。2026-10-20 より前は履歴から機械で遡った値）
- source-licenses.csv: 出典のライセンスの記録（2026-10-20 に OpenAlex で確かめた分。NC・ND の出典は、根拠の文に逐語の引用を入れず言い換えている）
- entity-media.csv / entity-figures.csv: 写真と模式図の台帳（画像本体は https://nigamilab.com/media/ 以下）

作り方・確度ラベル・確認状態の定義: https://nigamilab.com/protocol/
訂正の連絡: https://nigamilab.com/about/

----------------------------------------------------------------------
NIGAMI LAB dataset (English)

All files are UTF-8 CSV with a header row; rows reference each other by id.
- entities.csv: entries (ingredients, organisms, compounds, receptors, sensations, processes, dishes, seasonings, spices, regions, cultures, concepts)
- names.csv: names of entries (language, script, kind; rows with script=Hira whose notes start with 読み仮名 are sort keys)
- sources.csv: source ledger (review_status = how far each source was checked: full_text_checked / abstract_checked / metadata_only)
- claims.csv: claims (subject -> predicate -> object; claim_status = confidence: established / supported / preliminary / contested / contextual / observation / unverified)
- evidence.csv: links between claims and sources (relation: supports / contradicts / limits / contextualizes; finding_paraphrase summarises the source, limitations qualifies it)
- entity-overviews.csv / entity-overview-evidence.csv: prose paragraphs per entry and the evidence each paragraph used
- measurements.csv: thresholds, concentrations and other numbers
- recipes.csv: composition of dishes (verification_status)
- leads.csv: hypothesis box (questions kept as questions)
- source-checks.csv: record of how each source was checked (when and how far; scope = 全文 full / 部分 partial / 抄録 abstract; rows before 2026-10-20 were traced back by machine from the ledger notes and history)
- source-licenses.csv: license record of sources (checked via OpenAlex on 2026-10-20; for NC or ND sources, the evidence text paraphrases instead of quoting verbatim)
- entity-media.csv / entity-figures.csv: ledgers of photographs and diagrams (image files: https://nigamilab.com/media/)

The edited and structured data is CC BY 4.0. Verbatim quotations from third-party papers, dictionaries, standards and official pages
(inside 『 』 etc.) and the image files themselves keep their original rights; see LICENSE.txt.
Method, confidence labels and review states: https://nigamilab.com/en/protocol/
Corrections: https://nigamilab.com/en/about/
The text of the site is Japanese; an English edition of the topic hubs and research notes is at https://nigamilab.com/en/
