Medically Aware GPT-3 as a Data Generator for Medical Dialogue Summarization

TL;DR

This paper introduces a GPT-3 based method to generate synthetic medical dialogue summaries, significantly reducing the need for large labeled datasets while maintaining high medical accuracy and coherence.

Contribution

The authors develop an algorithm using GPT-3 to create high-quality synthetic training data for medical dialogue summarization, outperforming traditional data requirements.

Findings

Synthetic data achieves comparable performance to extensive human-labeled data.

Combining synthetic and human data improves summary quality.

The approach enhances medical accuracy and coherence in summaries.

Abstract

In medical dialogue summarization, summaries must be coherent and must capture all the medically relevant information in the dialogue. However, learning effective models for summarization require large amounts of labeled data which is especially hard to obtain. We present an algorithm to create synthetic training data with an explicit focus on capturing medically relevant information. We utilize GPT-3 as the backbone of our algorithm and scale 210 human labeled examples to yield results comparable to using 6400 human labeled examples (~30x) leveraging low-shot learning and an ensemble method. In detailed experiments, we show that this approach produces high quality training data that can further be combined with human labeled data to get summaries that are strongly preferable to those produced by models trained on human data alone both in terms of medical accuracy and coherency.