Abstract

NoteAid is a BioNLP system that automatically recognizes medical concepts and translates medical jargon into patient-oriented lay language. The lay language definition are annotated by professional human annotators, but the process is expensive and time-consuming.

This project is an attempt to publish a dataset comprised of human-annotated lay definitions and 3 external knowledge resources including the Unified Medical Language System (UMLS), Wikipedia, and Dictionary by Merriam-Webster. Also, we use transformer-based architectures to train and generate Lay Definitions with given jargon term and definitions from external knowledge resources.

Keywords: Electronic Health Records, Consumer Health, Information Retrieval, Natural Language Processing

Claim

My work is conducted by the time I was working in UMass BioNLP Lab. Now I have left as I graduated from UMass Amherst. However, the research is still actively work in progress led by Zonghai Yao.