Asclepius: Synthetic Clincal Notes & Instruction Dataset

Dataset Description

• Repository: Github
• Paper: https://arxiv.org/abs/2309.00237

Dataset Summary

This dataset is official dataset for Asclepius (arxiv) This dataset is composed with Clinical Note - Question - Answer format to build a clinical LLMs.

• We first synthesized synthetic notes from PMC-Patients case reports with GPT-3.5
• Then, we generate instruction-answer pairs for 157k synthetic discharge summaries

Supported Tasks

• This dataset covers below 8 tasks
  • Named Entity Recognition
  • Abbreviation Expansion
  • Relation Extraction
  • Temporal Information Extraction
  • Coreference Resolution
  • Paraphrasing
  • Summarization
  • Question Answering

Languages

English

Dataset Structure

Data Instances

• synthetic.csv
  • Clinical Note - Question - Answer pairs

Data Fields

• patient_id: Unique case report id from PMC-Patients
• patient: Case report text
• question: GPT-3.5 generated instruction from patient. The used prompt can be checked on github.
• answer: GPT-3.5 generated answer for given case report and question
• task: Corresponding category of question. One of above listsed

Dataset Creation

Source Data

PMC-Patients

Annotations

We used GPT-3.5-turbo (version 0314). You can check the prompts on our github.

Additional Information

Models

• Asclepius-7B
• Asclepius-13B
• Asclepius-Llama2-7B
• Asclepius-Llama2-13B
• Asclepius-Llama3-8B
• Asclepius-Mistral-7B-v0.3

Variants

• The instruction-answer pairs generated from MIMIC-III discharge summaries and the models trained with them are now available on Physionet!

Licensing Information

CC-BY-NC-SA 4.0

Citation Information

@misc{kweon2023publicly,
    title={Publicly Shareable Clinical Large Language Model Built on Synthetic Clinical Notes},
    author={Sunjun Kweon and Junu Kim and Jiyoun Kim and Sujeong Im and Eunbyeol Cho and Seongsu Bae and Jungwoo Oh and Gyubok Lee and Jong Hak Moon and Seng Chan You and Seungjin Baek and Chang Hoon Han and Yoon Bin Jung and Yohan Jo and Edward Choi},
    year={2023},
    eprint={2309.00237},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}