LA-RAG:Enhancing LLM-based ASR Accuracy with Retrieval-Augmented Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Shaojun, Shang, Hengchao, Wei, Daimeng, Guo, Jiaxin, Li, Zongyao, He, Xianghui, Zhang, Min, Yang, Hao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909314476670976
author Li, Shaojun
Shang, Hengchao
Wei, Daimeng
Guo, Jiaxin
Li, Zongyao
He, Xianghui
Zhang, Min
Yang, Hao
author_facet Li, Shaojun
Shang, Hengchao
Wei, Daimeng
Guo, Jiaxin
Li, Zongyao
He, Xianghui
Zhang, Min
Yang, Hao
contents Recent advancements in integrating speech information into large language models (LLMs) have significantly improved automatic speech recognition (ASR) accuracy. However, existing methods often constrained by the capabilities of the speech encoders under varied acoustic conditions, such as accents. To address this, we propose LA-RAG, a novel Retrieval-Augmented Generation (RAG) paradigm for LLM-based ASR. LA-RAG leverages fine-grained token-level speech datastores and a speech-to-speech retrieval mechanism to enhance ASR accuracy via LLM in-context learning (ICL) capabilities. Experiments on Mandarin and various Chinese dialect datasets demonstrate significant improvements in ASR accuracy compared to existing methods, validating the effectiveness of our approach, especially in handling accent variations.
format Preprint
id arxiv_https___arxiv_org_abs_2409_08597
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LA-RAG:Enhancing LLM-based ASR Accuracy with Retrieval-Augmented Generation
Li, Shaojun
Shang, Hengchao
Wei, Daimeng
Guo, Jiaxin
Li, Zongyao
He, Xianghui
Zhang, Min
Yang, Hao
Sound
Computation and Language
Audio and Speech Processing
Recent advancements in integrating speech information into large language models (LLMs) have significantly improved automatic speech recognition (ASR) accuracy. However, existing methods often constrained by the capabilities of the speech encoders under varied acoustic conditions, such as accents. To address this, we propose LA-RAG, a novel Retrieval-Augmented Generation (RAG) paradigm for LLM-based ASR. LA-RAG leverages fine-grained token-level speech datastores and a speech-to-speech retrieval mechanism to enhance ASR accuracy via LLM in-context learning (ICL) capabilities. Experiments on Mandarin and various Chinese dialect datasets demonstrate significant improvements in ASR accuracy compared to existing methods, validating the effectiveness of our approach, especially in handling accent variations.
title LA-RAG:Enhancing LLM-based ASR Accuracy with Retrieval-Augmented Generation
topic Sound
Computation and Language
Audio and Speech Processing
url https://arxiv.org/abs/2409.08597