Synthetic Data Augmentation Enhance Disease Named Entity Recognition

Poster Number: P105
Presentation Time: 05:00 PM - 06:30 PM

Abstract Keywords: Information Extraction, Natural Language Processing, Large Language Models (LLMs), Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics

Synthetically generated labelled examples from language models have the potential address coverage gaps and imbalances in training corpora. We use Unified Medical Language System (UMLS) sourced prompts in conjunction with ChatGPT model to generate clinical mentions of diseases and show a small but significant improvement when including synthetic text.

Speaker(s):
John Osborne, PhD
University of Alabama at Birmingham

Author(s):
Kuleen Sasse;