Task-Specific Transformer-Based Language Models in Health Care: Scoping Review (Preprint)

Abstract

BACKGROUND In the field of artificial intelligence, language models, which are used to convey knowledge in the medical domain, have rapidly increased in number. However, no comprehensive review is available to guide researchers in constructing and applying language models for medical applications. OBJECTIVE We aim to leverage the power of these language models to improve healthcare by addressing the challenges in the six tasks we reviewed. METHODS We present potential solutions to the identified limitations to provide useful insights for future research in natural language processing and the development of language models for medical applications. RESULTS We surveyed studies on medical transformer-based language models, categorizing them into six tasks: dialogue generation, question-answering, summarization, text classification, sentiment analysis, and named entity recognition. CONCLUSIONS By proposing potential solutions, we hope to facilitate the creation of more effective and accurate language models that can be utilized to enhance healthcare delivery and improve patient outcomes.