Retrieval Augmented Correction of Named Entity Speech Recognition Errors

TL;DR

This work proposes a RAG-like technique for correcting speech recognition entity name errors which uses a vector database to index a set of relevant entities and achieves 33%-39% relative word error rate reductions on synthetic test sets.

Abstract

In recent years, end-to-end automatic speech recognition (ASR) systems have proven themselves remarkably accurate and performant, but these systems still have a significant error rate for entity names which appear infrequently in their training data. In parallel to the rise of end-to-end ASR systems, large language models (LLMs) have proven to be a versatile tool for various natural language processing (NLP) tasks. In NLP tasks where a database of relevant knowledge is available, retrieval augmented generation (RAG) has achieved impressive results when used with LLMs. In this work, we propose a RAG-like technique for correcting speech recognition entity name errors. Our approach uses a vector database to index a set of relevant entities. At runtime, database queries are generated from possibly errorful textual ASR hypotheses, and the entities retrieved using these queries are fed, along with the ASR hypotheses, to an LLM which has been adapted to correct ASR errors. Overall, our best system achieves 33%-39% relative word error rate reductions on synthetic test sets focused on voice assistant queries of rare music entities without regressing on the STOP test set, a publicly available voice assistant test set covering many domains.

Figures (1)

• Figure 1: Our RAG-based approach. with an example. ASR errors are in red and corrected errors are in green. In the context, [H] precedes each hint, [A] precedes the ASR transcript and [P] is the cue for the LLM to start prediction.