A Comprehensive Solution to Connect Speech Encoder and Large Language Model for ASR

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Pham, Van Tung, Lin, Yist, Han, Tao, Li, Wei, Zhang, Jun, Lu, Lu, Wang, Yuxuan
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910500258840576
author Pham, Van Tung
Lin, Yist
Han, Tao
Li, Wei
Zhang, Jun
Lu, Lu
Wang, Yuxuan
author_facet Pham, Van Tung
Lin, Yist
Han, Tao
Li, Wei
Zhang, Jun
Lu, Lu
Wang, Yuxuan
contents Recent works have shown promising results in connecting speech encoders to large language models (LLMs) for speech recognition. However, several limitations persist, including limited fine-tuning options, a lack of mechanisms to enforce speech-text alignment, and high insertion errors especially in domain mismatch conditions. This paper presents a comprehensive solution to address these issues. We begin by investigating more thoughtful fine-tuning schemes. Next, we propose a matching loss to enhance alignment between modalities. Finally, we explore training and inference methods to mitigate high insertion errors. Experimental results on the Librispeech corpus demonstrate that partially fine-tuning the encoder and LLM using parameter-efficient methods, such as LoRA, is the most cost-effective approach. Additionally, the matching loss improves modality alignment, enhancing performance. The proposed training and inference methods significantly reduce insertion errors.
format Preprint
id arxiv_https___arxiv_org_abs_2406_17272
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle A Comprehensive Solution to Connect Speech Encoder and Large Language Model for ASR
Pham, Van Tung
Lin, Yist
Han, Tao
Li, Wei
Zhang, Jun
Lu, Lu
Wang, Yuxuan
Machine Learning
Recent works have shown promising results in connecting speech encoders to large language models (LLMs) for speech recognition. However, several limitations persist, including limited fine-tuning options, a lack of mechanisms to enforce speech-text alignment, and high insertion errors especially in domain mismatch conditions. This paper presents a comprehensive solution to address these issues. We begin by investigating more thoughtful fine-tuning schemes. Next, we propose a matching loss to enhance alignment between modalities. Finally, we explore training and inference methods to mitigate high insertion errors. Experimental results on the Librispeech corpus demonstrate that partially fine-tuning the encoder and LLM using parameter-efficient methods, such as LoRA, is the most cost-effective approach. Additionally, the matching loss improves modality alignment, enhancing performance. The proposed training and inference methods significantly reduce insertion errors.
title A Comprehensive Solution to Connect Speech Encoder and Large Language Model for ASR
topic Machine Learning
url https://arxiv.org/abs/2406.17272