Adapting Large Language Model with Speech for Fully Formatted End-to-End Speech Recognition
Authors:
- Shaoshi LingSLShaoshi Ling
- Yuxuan HuYHYuxuan Hu
- Shuangbei QianSQShuangbei Qian
- Guoli YeGYGuoli Ye
- Yao QianYQYao Qian
- Yifan GongYGYifan Gong
- Ed LinELEd Lin
- Michael ZengMZMichael Zeng
Preprint
Abstract
Most end-to-end (E2E) speech recognition models are composed of encoder and decoder blocks that perform acoustic and language modeling functions. Pretrained large language models (LLMs) have the potential to improve the performance of E2E ASR. However, integrating a pretrained language model into an E2E speech recognition model has shown limited benefits due to the mismatches between text-based LLMs and those used in E2E ASR. In this paper, we explore an alternative approach by adapting a pretrained LLMs to speech. Our experiments on fully-formatted E2E ASR transcription tasks across various domains demonstrate that our approach can effectively leverage the strengths of pretrained LLMs to produce more readable ASR transcriptions. Our model, which is based on the pretrained large language models with either an encoder-decoder or decoder-only structure, surpasses strong ASR models such as Whisper, in terms of recognition error rate, considering formats like punctuation and capitalization as well.Keywords
All Related Versions
This publication has 7 references indexed in Scilit:
- Improving Readability for Automatic Speech Recognition TranscriptionACM Transactions on Asian and Low-Resource Language Information Processing, 2023
- MAESTRO: Matched Speech Text Representations through Modality Matching2022
- Unified Speech-Text Pre-training for Speech Translation and Recognition2022
- FSR: Accelerating the Inference Process of Transducer-Based Models by Applying Fast-Skip Regularization2021
- Multilingual Speech Translation from Efficient Finetuning of Pretrained Models2021
- Cold Fusion: Training Seq2Seq Models Together with Language Models2018
- Towards Better Decoding and Language Model Integration in Sequence to Sequence Models2017