Alignment Entropy Regularization

TL;DR

This paper introduces entropy regularization in ASR training to reduce alignment uncertainty, resulting in simpler decoding and improved alignment quality without increasing word error rate.

Contribution

It proposes using entropy as a regularizer to constrain alignment distributions, enhancing decoding simplicity and alignment accuracy in speech recognition.

Findings

Simpler decoding without WER increase

Improved time alignment quality

Entropy regularization reduces alignment uncertainty

Abstract

Existing training criteria in automatic speech recognition(ASR) permit the model to freely explore more than one time alignments between the feature and label sequences. In this paper, we use entropy to measure a model's uncertainty, i.e. how it chooses to distribute the probability mass over the set of allowed alignments. Furthermore, we evaluate the effect of entropy regularization in encouraging the model to distribute the probability mass only on a smaller subset of allowed alignments. Experiments show that entropy regularization enables a much simpler decoding method without sacrificing word error rate, and provides better time alignment quality.