DeRAGEC: Denoising Named Entity Candidates with Synthetic Rationale for ASR Error Correction

TL;DR

DeRAGEC enhances NE correction in ASR by filtering noisy candidates with synthetic rationales, leveraging phonetic similarity and in-context learning, leading to significant WER improvements without extra training.

Contribution

It introduces DeRAGEC, a novel method that uses synthetic denoising rationales to improve NE correction in ASR systems without additional training.

Findings

28% relative WER reduction on datasets

Outperforms baseline ASR and RAGEC methods

Effective filtering of noisy NE candidates

Abstract

We present DeRAGEC, a method for improving Named Entity (NE) correction in Automatic Speech Recognition (ASR) systems. By extending the Retrieval-Augmented Generative Error Correction (RAGEC) framework, DeRAGEC employs synthetic denoising rationales to filter out noisy NE candidates before correction. By leveraging phonetic similarity and augmented definitions, it refines noisy retrieved NEs using in-context learning, requiring no additional training. Experimental results on CommonVoice and STOP datasets show significant improvements in Word Error Rate (WER) and NE hit ratio, outperforming baseline ASR and RAGEC methods. Specifically, we achieved a 28% relative reduction in WER compared to ASR without postprocessing. Our source code is publicly available at: https://github.com/solee0022/deragec