seiya/jp-disease-finding-dataset

(日本語版はこちら)

Overview

This dataset provides information extracted from ≈ 7 k Japanese medical-journal articles (The Journal of the Japanese Society of Internal Medicine, 2003 – 2023).

For each row we include:

• disease_text – the disease name (Japanese)
• findings – a list of related symptoms / examinations / complications
• finding_description – sentences quoted verbatim to show how the term is used
• article metadata – article_id, authors, journal_meta, url

The JSON-Lines file disease_level.jsonl is one row per disease.

---

Licensing & reuse

Part	File / column	Licence	Permitted uses
Structured facts	disease_level.jsonl (everything except finding_description)	CC BY 4.0	Commercial and non-commercial redistribution, adaptation and machine-learning permitted with attribution.
Quoted Japanese snippets	finding_description column	journal copyright policy (see below)	non-commercial use recommended; no alteration of wording; keep article metadata

Journal copyright policy

The Japanese-language journal does not display an explicit Creative Commons licence.
Its sister English journal Internal Medicine states ”CC BY-NC-ND 4.0”:

本誌はオープンアクセスジャーナルであり、クリエイティブ・コモンズ 表示 - 非営利 - 改変禁止 4.0 国際 ライセンスの下に公開している。
(https://internmed.jp/modules/authors_jp/index.php?content_id=1)

We therefore treat each quotation as a lawful academic quotation.

OpenAI output notice

Portions of this dataset were produced with OpenAI GPT-4 (API version gpt-4.1-mini-2025-04-14).

Under the OpenAI Terms of Use (December 11, 2024) § Prohibited Conduct you may not use these outputs to develop models that compete with OpenAI. (https://openai.com/policies/row-terms-of-use/)

---

Acknowledgment

This work was supported by JSPS KAKENHI Grant Number JP22K12263.

Citation

If you use this dataset, please cite:
Currently under preparation...