Unsupervised Data Selection via Discrete Speech Representation for ASR

Self-supervised learning of speech representations has achieved impressive results in improving automatic speech recognition (ASR). In this paper, we show that data selection is important for self-supervised learning. We propose a simple and effective unsupervised data selection method which selects...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2022-04
Main Authors: Lu, Zhiyun, Wang, Yongqiang, Zhang, Yu, Han, Wei, Chen, Zhehuai, Haghani, Parisa
Format: Article
Language:English
Subjects:
Automatic speech recognition
Error reduction
Representations
Supervised learning
Training
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Self-supervised learning of speech representations has achieved impressive results in improving automatic speech recognition (ASR). In this paper, we show that data selection is important for self-supervised learning. We propose a simple and effective unsupervised data selection method which selects acoustically similar speech to a target domain. It takes the discrete speech representation available in common self-supervised learning frameworks as input, and applies a contrastive data selection method on the discrete tokens. Through extensive empirical studies we show that our proposed method reduces the amount of required pre-training data and improves the downstream ASR performance. Pre-training on a selected subset of 6% of the general data pool results in 11.8% relative improvements in LibriSpeech test-other compared to pre-training on the full set. On Multilingual LibriSpeech French, German, and Spanish test sets, selecting 6% data for pre-training reduces word error rate by more than 15% relatively compared to the full set, and achieves competitive results compared to current state-of-the-art performances.
ISSN:2331-8422