Supervised Attention in Sequence-to-Sequence Models for Speech Recognition

TL;DR

This paper introduces a supervised attention mechanism in sequence-to-sequence speech recognition models, improving alignment accuracy and overall performance by explicitly guiding the attention weights during training.

Contribution

It proposes a novel supervised attention loss that enhances alignment learning in speech recognition models, addressing the gap between attention weights and true alignments.

Findings

Significant performance improvements with supervised attention

Better alignment accuracy in speech recognition models

Supervised attention crucial for model performance

Abstract

Attention mechanism in sequence-to-sequence models is designed to model the alignments between acoustic features and output tokens in speech recognition. However, attention weights produced by models trained end to end do not always correspond well with actual alignments, and several studies have further argued that attention weights might not even correspond well with the relevance attribution of frames. Regardless, visual similarity between attention weights and alignments is widely used during training as an indicator of the models quality. In this paper, we treat the correspondence between attention weights and alignments as a learning problem by imposing a supervised attention loss. Experiments have shown significant improved performance, suggesting that learning the alignments well during training critically determines the performance of sequence-to-sequence models.