EXPLORING MACHINE SPEECH CHAIN FOR DOMAIN ADAPTATION

Yue，Fengpeng1; Deng，Yan2; He，Lei2; Ko，Tom1; Zhang，Yu1,3

10.1109/ICASSP43922.2022.9746721

2022

47th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

1520-6149

978-1-6654-0541-6

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

2022-May

6757-6761

23-27 May 2022

Singapore, Singapore

345 E 47TH ST, NEW YORK, NY 10017 USA

IEEE

Machine Speech Chain integrates both end-to-end (E2E) automatic speech recognition (ASR) and neural text-to-speech (TTS) into one circle for joint training. It has been proven that it can effectively leverage a large amount of unpaired data in the spirit of data augmentation. In this paper, we explore the TTS→ASR pipeline in machine speech chain to perform domain adaptation for both E2E ASR and neural TTS models with only text data from the target domain. We conduct experiments by adapting from audiobook domain (i.e., LibriSpeech) to presentation domain (i.e., TED-LIUM). There is a relative word error rate (WER) reduction of 19.7% for the E2E ASR model on the TED-LIUM test set, and a relative WER reduction of 29.4% in synthetic speech generated by neural TTS in the presentation domain. Moreover, we observe that the gains from the proposed method and conventional adaptation methods of language models are additive.

domain adaptation machine speech chain speech recognition speech synthesis

第一 ; 通讯

英语

EI ; CPCI-S

Acoustics ; Computer Science ; Engineering

Acoustics ; Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic

WOS:000864187907012

20222312198862

2-s2.0-85131246775

Scopus

1.Department of Computer Science and Engineering,Southern University of Science and Technology,Shenzhen,China
2.Microsoft China,China
3.Peng Cheng Laboratory,Shenzhen,China

Yue，Fengpeng,Deng，Yan,He，Lei,et al. EXPLORING MACHINE SPEECH CHAIN FOR DOMAIN ADAPTATION[C]. 345 E 47TH ST, NEW YORK, NY 10017 USA:IEEE,2022:6757-6761.