PULSAR at MEDIQA-Sum 2023: Large Language Models Augmented by Synthetic Dialogue Convert Patient Dialogues to Medical Records

Viktor Schlegel, Hao Li, Yuping Wu, Anand Subramanian, Thanh-Tung Nguyen +6 more
7/5/2023
cs.CL

Official code: yuping-wu/PULSAR (6 stars, last commit 2023-07-09).

Abstract

This paper describes PULSAR, our system submission at the ImageClef 2023 MediQA-Sum task on summarising patient-doctor dialogues into clinical records. The proposed framework relies on domain-specific pre-training, to produce a specialised language model which is trained on task-specific natural data augmented by synthetic data generated by a black-box LLM. We find limited evidence towards the efficacy of domain-specific pre-training and data augmentation, while scaling up the language model yields the best performance gains. Our approach was ranked second and third among 13 submissions on task B of the challenge. Our code is available at https://github.com/yuping-wu/PULSAR.

View on arXivView PDF

Code Implementations(1)

yuping-wu/PULSAROfficial100%

PULSAR: Pre-training with Extracted Gap Healthcare Terms for Summarizing Patients’ Problems and Data Augmentation with Black-box Large Language Models

61Shell, PythonMar 8, 20233 years ago

Similar Papers

ClinicalGPT: Large Language Models Finetuned with Diverse Medical Data and Comprehensive Evaluation

85% match2023

PMC-LLaMA: Towards Building Open-source Language Models for Medicine

84% match1 repo6802023

Do We Still Need Clinical Language Models?

83% match1 repo2023

Instruction Tuning Large Language Models to Understand Electronic Health Records

83% match1 repo612024

Clinical Camel: An Open Expert-Level Medical Language Model with Dialogue-Based Knowledge Encoding

83% match1 repo542023

MedChatZH: a Better Medical Adviser Learns from Better Instructions

83% match1 repo1102023

Cite this paper

@article{schlegel2023pulsar,
  title  = {PULSAR at MEDIQA-Sum 2023: Large Language Models Augmented by Synthetic Dialogue Convert Patient Dialogues to Medical Records},
  author = {Viktor Schlegel and Hao Li and Yuping Wu and Anand Subramanian and Thanh-Tung Nguyen and Abhinav Ramesh Kashyap and Daniel Beck and Xiaojun Zeng and Riza Theresa Batista-Navarro and Stefan Winkler and Goran Nenadic},
  year   = {2023},
  eprint = {2307.02006},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2307.02006v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2307.02006.svg)](https://codeofpaper.com/paper/2307.02006)

Discussion