---
license: apache-2.0
---
# # Medbase_data: Multilingual Medical LLM Dataset (EN/ZH/ES/FR/AR/HI)

<p align="center">
   📃 <a href="" target="_blank">Paper</a> • 🌐 <a href="https://github.com/FreedomIntelligence/Medbase" target="_blank">Github</a> • 🤗 <a href="https://huggingface.co/datasets/FreedomIntelligence/Medbase_data" target="_blank">HuggingFace</a>  
   <br> 
</p>



![Medbase](assets/Medbase.png)

## Usage

- [Zip File](https://huggingface.co/datasets/FreedomIntelligence/Medbase_data/blob/main/Medbase_data-datasets.zip)
- [Data category](https://huggingface.co/datasets/FreedomIntelligence/Medbase_data/tree/main/train)
  - Pretrain:
    - json_name: {data_source}_{language}_{data_type}.json
      - data_type: medicalBook, medicalGuideline, medicalPaper, medicalWeb(from online forum), medicalWiki
      - language: en(English), zh(chinese), es(spanish), fr(french), hi(Hindi)
      - data_type: qa(generated qa from text)
    - data item:
      - data_type==text: list of string
        ```
        [
          "string1",
          "string2",
          ...
        ]
        ```
      - data_type==qa: list of qa pairs(list of string)
        ```
        [
          [
            "q1",
            "a1",
            "q2",
            "a2",
            ...
          ],
          ...
        ]
        ```

  - SFT:
    - json_name: {data_source}_{language}.json
      - data_type: code, general, math, medicalExam, medicalPatient
    - data item: list of qa pairs(list of string)
      ```
        [
          [
            "q1",
            "a1",
            "q2",
            "a2",
            ...
          ],
          ...
        ]
        ```

    

## **Data:** Huge, Diverse, Clean, Multilingual

| Data Type          | Description                  | Source(ZH)                                                   | Source(EN)                                                   | Source(FR)                                                   | Source(ES)                                               | Source(AR)                                                   | Source(HI)                                                   |
| ------------------ | ---------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | -------------------------------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ |
| Continue Pretrain  |                              |                                                              |                                                              |                                                              |                                                          |                                                              |                                                              |
| Medical Books      | Medical related Books        | MedQA-books                                                  | Pile-Books                                                   | -                                                            |                                                          |                                                              |                                                              |
| Medical Guidelines | Clinical Medicine Guide      | Chinese Medical Association                                  | [Medtron guideline](https://huggingface.co/datasets/epfl-llm/guidelines) | -                                                            |                                                          |                                                              |                                                              |
| Medical Wiki       | Medical related wikipedia    | Wikipedia & Wikidoc                                          | Wikipedia  & Wikidoc                                         | [CLEAR - Simple Corpus for Medical French](http://natalia.grabar.free.fr/resources.php#clear) | -                                                        | -                                                            | [Hindi_health](https://www.kaggle.com/datasets/aijain/hindi-health-dataset/data?select=Symptom+Gazetteer.txt) |
| Medical Paper      | Medical related paper        | Papers abstract                                              | PubMed Abstract                                              | [MORFITT](https://huggingface.co/datasets/qanastek/MORFITT?row=98): Pubmed-french Cochrane: [CLEAR-](http://natalia.grabar.free.fr/resources.php#clear)abs | [Mesinesp](https://zenodo.org/records/3826492)           | -                                                            | -                                                            |
| Medical Web        | Medical related web data     | Wudao                                                        | C4                                                           | [Frenchmedmcqa](https://github.com/qanastek/FrenchMedMCQA)_train | [CoWeSe](https://zenodo.org/records/5513237)             | -                                                            | -                                                            |
| SFT                |                              |                                                              |                                                              |                                                              |                                                          |                                                              |                                                              |
| Medical Exam       | Medical related exams        | MedQA CExam CMB (Train Set)                                  | MedQA MedmcQA PubMedQA  (Train Set)                          | -                                                            | [Head_qa](https://huggingface.co/datasets/head_qa)_train | -                                                            | -                                                            |
| Medical Patient    | Doctor-patient dialogue data | [HuatuoGPT-I](https://huggingface.co/datasets/FreedomIntelligence/HuatuoGPT-sft-data-v1) | [PMC_patients](https://huggingface.co/datasets/zhengyun21/PMC-Patients?row=34) | -                                                            | -                                                        | [MAQA](https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/Y2JBEZ) | -                                                            |
| General_Replay     | General SFT Data             | Wizard & ShareGPT & Alpaca                                   | Wizard & ShareGPT & Alpaca & [Dataset List](https://huggingface.co/jondurbin/bagel-dpo-34b-v0.2#sft-data-sources) | ShareGPT & Alpaca                                            | ShareGPT & Alpaca                                        | ShareGPT & Alpaca                                            | ShareGPT & Alpaca                                            |
| Code               | Code Data                    | [leetcode-11k](https://huggingface.co/datasets/krisfu/awesome-llm-datasets-only-Chinese) | [python_alpaca](https://huggingface.co/datasets/Vezora/Tested-22k-Python-Alpaca) | -                                                            | -                                                        | -                                                            | -                                                            |
| Math               | Math Data                    |                                                              | [mathinstruct](https://huggingface.co/datasets/TIGER-Lab/MathInstruct) | -                                                            | -                                                        | -                                                            | -                                                            |


## Citation

```
@misc{medbase,
  title={MedBase, Exploring the boundaries of open source LLM medical capabilities},
  author={Xidong Wang, Junyin Chen, Nuo Chen, Yidong Wang, Zhiyi Zhang, Benyou Wang},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/FreedomIntelligence/Medbase}},
}
```