Skip to content

believewhat/Dr.NoteAid

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

52 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Can LLMs generate high-quality synthetic note-oriented doctor-patient conversations?

Image text

Introduction

In the MEDIQA-Chat 2023, we focus especially on Task-C and propose a novel LLMs cooperation system named a doctor-patient loop to generate high-quality conversation data sets. The experiment results demonstrate that our approaches yield excellent performance as evaluated by automatic metrics such as ROUGE, medical concept recall, BLEU, and Self-BLEU. We won second place in the competition, but after further prompt engineering, our method can finally achieve better results than the first place in the competition. Furthermore, we conducted a comparative analysis between our proposed method and ChatGPT and GPT-4, thus providing further clarification on why our approach yields superior results. This analysis also investigates the potential of utilizing cooperation LLMs to generate high-quality datasets. Please see our paper for more details: xxxxxxxx.

Datsets Links (10k version):

  1. ChatGPT Dataset
  2. GPT-4 Dataset
  3. NoteChat Dataset

We have now uploaded all versions of the datasets to Hugging Face: link

We update our codes which is based on gpt-4 module version. Some prompt details or parameters should be adjusted based on different task or note.

Plan

  • Expand the ChatGPT dataset(about 167k).
  • Expand the GPT-4 dataset.
  • Expand our NoteChat-ChatGPT dataset.
  • Expand our NoteChat-GPT4 dataset.
  • Finetune chatbot models.

LLM Model

  1. Demo
  2. Model Link
@article{wang2023notechat,
  title={NoteChat: A Dataset of Synthetic Doctor-Patient Conversations Conditioned on Clinical Notes},
  author={Wang, Junda and Yao, Zonghai and Yang, Zhichao and Zhou, Huixue and Li, Rumeng and Wang, Xun and Xu, Yucheng and Yu, Hong},
  journal={arXiv preprint arXiv:2310.15959},
  year={2023}
}

@inproceedings{wang-etal-2023-umass,
    title = "{UMASS}{\_}{B}io{NLP} at {MEDIQA}-Chat 2023: Can {LLM}s generate high-quality synthetic note-oriented doctor-patient conversations?",
    author = "Wang, Junda  and
      Yao, Zonghai  and
      Mitra, Avijit  and
      Osebe, Samuel  and
      Yang, Zhichao  and
      Yu, Hong",
    editor = "Naumann, Tristan  and
      Ben Abacha, Asma  and
      Bethard, Steven  and
      Roberts, Kirk  and
      Rumshisky, Anna",
    booktitle = "Proceedings of the 5th Clinical Natural Language Processing Workshop",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.clinicalnlp-1.49",
    doi = "10.18653/v1/2023.clinicalnlp-1.49",
    pages = "460--471",
    abstract = "This paper presents UMASS{\_}BioNLP team participation in the MEDIQA-Chat 2023 shared task for Task-A and Task-C. We focus especially on Task-C and propose a novel LLMs cooperation system named a doctor-patient loop to generate high-quality conversation data sets. The experiment results demonstrate that our approaches yield reasonable performance as evaluated by automatic metrics such as ROUGE, medical concept recall, BLEU, and Self-BLEU. Furthermore, we conducted a comparative analysis between our proposed method and ChatGPT and GPT-4. This analysis also investigates the potential of utilizing cooperation LLMs to generate high-quality datasets.",
}

About

ACL Workshop 2023

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages