Sofun2009/yzint-drug-data

pretty_name: CN-Drug-KG-800
language:
  - zh
license: cc-by-nc-4.0
tags:
  - medical
  - knowledge-graph
  - information-extraction
  - chinese
task_categories:
  - information-extraction
  - text-retrieval
  - other
size_categories:
  - 1K<n<100M
multilinguality: monolingual

---

CN-Drug-KG-800: Chinese Drug Knowledge Graph Dataset

One-line summary: A Chinese-language drug knowledge graph dataset containing about 800 drug entities, along with their semantic relations to ingredients, administration routes, and more. Suitable for information extraction, knowledge graph construction, and retrieval-augmented generation (RAG).

🧭 Dataset Summary

• Language: Chinese (Simplified)

• Entities: ~800 drug entities, related ingredients, administration routes, etc.

• Relation types: Curated from drug instruction semantics, including “composed of”, “used for/indication”, “administration route”, “caution/contraindication”, “interaction”, etc. (example below shows administration route)

• Data format: Two types of JSON/JSONL files:

  • entities: entity table (fields: code, name, type, description)
  • relations: relation table (fields: source, sourceName, target, targetName, relationship)

• Typical use cases: drug knowledge retrieval, RAG knowledge bases, KG construction/visualization, entity & relation extraction, Chinese biomedical NLP

• Disclaimer: For research and educational purposes only. Not medical advice. Not for clinical diagnosis or treatment decision-making.

📦 Repository Layout

.
├── data/
│   ├── entities.json        # or entities.jsonl
│   └── relations.json       # or relations.jsonl
├── README.md
└── LICENSE

🧱 Dataset Structure

Entities

Field	Type	Description
code	int/str	Unique entity identifier (numeric or long integer convertible to string)
name	str	Entity name (e.g., “罗汉果止咳膏”, “桑白皮”)
type	str	Entity type (e.g., drug, ingredient; can extend to route, indication, etc.)
description	str	Natural language description (composition, indications, dosage, cautions, storage, etc.)

Example (excerpt):

{
  "code": 37487448644681728,
  "name": "罗汉果止咳膏",
  "type": "drug",
  "description": "Used for resolving phlegm and relieving cough, suitable for cold-induced cough and bronchitis. Main ingredients include Siraitia grosvenorii, loquat leaf, cortex mori, cynanchum, stemona, platycodon, menthol oil; excipients include sucrose, refined honey, tartaric acid. The product is a brown-yellow to dark brown viscous semi-fluid, sweet, slightly bitter, and cool in taste. Specifications: 200g/bottle and 250g/bottle. Dosage: oral, 10–15g per dose, three times daily. Contraindication: not for diabetic patients."
}

{
  "code": 37487448645009408,
  "name": "桑白皮",
  "type": "ingredient",
  "description": "An ingredient of Luo Han Guo Cough Syrup, with expectorant and antitussive functions."
}

Relations

Field	Type	Description
source	int/str	Source entity code
sourceName	str	Source entity name
target	int/str	Target entity code
targetName	str	Target entity name
relationship	str	Natural language description of relation between source and target

Example (excerpt):

{
  "sourceName": "肠内营养乳剂（TPF）",
  "targetName": "管饲或口服",
  "source": 37450838293352448,
  "target": 37450838293286912,
  "relationship": "Enteral nutrition emulsion (TPF) is administered via tube feeding or orally; dosage should be adjusted according to patient weight and nutritional status."
}

🔢 Dataset Statistics

• Drug entities: ~800
• Other entities: ingredients, administration routes, etc.
• Relations: aligned with entity coverage (grows with future versions)

🧪 Intended Uses

• Chinese biomedical NLP: NER, relation extraction (RE), event extraction (EE)
• Knowledge graph: drug knowledge integration, KG QA, graph retrieval/reasoning
• RAG applications: structured knowledge base (entity descriptions + relation statements)
• Teaching/Research: pharmacology education, information organization & retrieval

⚠️ Limitations & Ethical Considerations

• Not medical advice: Not for clinical decision-making. Use only in compliance with local regulations and ethics.
• Source variation: Drug information may vary by source and publication date. Always refer to the latest official instructions.
• Potential bias: Description length and terminology granularity may vary, affecting downstream consistency.
• Copyright compliance: Ensure redistribution rights if using public drug instructions/standards.

🏷️ Labels & Conventions

• type controlled vocabulary: drug, ingredient, route, indication, contraindication, interaction, etc.
• code is globally unique. To avoid long integer precision issues, convert to string when reading.

🔁 Versioning

• v1.0: Includes ~800 drug entities with example relations.
• Future updates: expand relation types, align with standard ontologies (ATC/ChEBI/MeSH mapping), add bilingual fields (Chinese/English).

📄 License

Released under CC BY-NC 4.0 (Attribution-NonCommercial). See LICENSE file for details.

🔍 Citation

If you use this dataset in your research or product, please cite:

@dataset{jieh_2025_cn_drug_kg_800,
  title   = {CN-Drug-KG-800: Chinese Drug Knowledge Graph Dataset},
  author  = {Chao Peng},
  year    = {2025},
  note    = {Chinese drug knowledge graph dataset with ~800 drug entities and relations}
}

---

Maintainer: Chao Peng (Insightful Intelligent) Contact: via issues/discussions in the dataset repository