MedSpeak: A Knowledge Graph-Aided ASR Error Correction Framework for Spoken Medical QA

TL;DR

MedSpeak is a novel framework that enhances spoken medical question-answering by correcting ASR errors using a medical knowledge graph and large language models, significantly improving accuracy.

Contribution

It introduces a knowledge graph-aided error correction method that leverages semantic and phonetic information, advancing medical spoken QA performance.

Findings

Significant improvement in medical term recognition accuracy.

Enhanced overall medical SQA performance.

Established as a state-of-the-art solution.

Abstract

Spoken question-answering (SQA) systems relying on automatic speech recognition (ASR) often struggle with accurately recognizing medical terminology. To this end, we propose MedSpeak, a novel knowledge graph-aided ASR error correction framework that refines noisy transcripts and improves downstream answer prediction by leveraging both semantic relationships and phonetic information encoded in a medical knowledge graph, together with the reasoning power of LLMs. Comprehensive experimental results on benchmarks demonstrate that MedSpeak significantly improves the accuracy of medical term recognition and overall medical SQA performance, establishing MedSpeak as a state-of-the-art solution for medical SQA. The code is available at https://github.com/RainieLLM/MedSpeak.