Advancing medical question answering with a knowledge embedding transformer

Abstract

Efficient medical question answering is essential for better patient care. Despite progress since Eliza (1966), even advanced LLMs (e.g., GPT-4) struggle with medical data. This study presents a system combining knowledge embedding and transformers. It includes a knowledge understanding layer and an answer generation layer. Tested on the MedQA dataset, it achieved 82.92% accuracy, outperforming GPT-4’s 71.07%. The results demonstrate the system’s ability to deliver accurate and ethical answers. This integrated method improves response speed and quality. Future work will enhance precision, support patient interaction, and integrate multimodal data for improved healthcare query processing.