Unsupervised data selection for Speech Recognition with contrastive loss ratios

TL;DR

This paper introduces an unsupervised data selection method for speech recognition using contrastive loss ratios and submodular functions, leading to improved WER performance over traditional likelihood-based methods.

Contribution

It presents a novel unsupervised data selection technique based on contrastive loss ratios and submodular optimization, enhancing speech recognition training efficiency and accuracy.

Findings

Outperforms likelihood-based selection in WER reduction

Reduces WER by 20.23% on Tedtalks dataset

Decreases WER by 6.26% on WSJCAM0 with less data

Abstract

This paper proposes an unsupervised data selection method by using a submodular function based on contrastive loss ratios of target and training data sets. A model using a contrastive loss function is trained on both sets. Then the ratio of frame-level losses for each model is used by a submodular function. By using the submodular function, a training set for automatic speech recognition matching the target data set is selected. Experiments show that models trained on the data sets selected by the proposed method outperform the selection method based on log-likelihoods produced by GMM-HMM models, in terms of word error rate (WER). When selecting a fixed amount, e.g. 10 hours of data, the difference between the results of two methods on Tedtalks was 20.23% WER relative. The method can also be used to select data with the aim of minimising negative transfer, while maintaining or improving on performance of models trained on the whole training set. Results show that the WER on the WSJCAM0 data set was reduced by 6.26% relative when selecting 85% from the whole data set.