Large scale evaluation of importance maps in automatic speech recognition

TL;DR

This paper introduces the structured saliency benchmark (SSBM) for evaluating importance maps in speech recognition, compares different importance map techniques, and finds bubble noise-based maps outperform energy-based baselines on a large dataset.

Contribution

It proposes a new evaluation metric for importance maps in speech recognition and conducts a large-scale comparison of different importance map generation techniques.

Findings

Bubble noise importance maps outperform energy-based maps

Evaluation metric applicable to structured prediction tasks

Large-scale comparison on 100 sentences from AMI corpus

Abstract

In this paper, we propose a metric that we call the structured saliency benchmark (SSBM) to evaluate importance maps computed for automatic speech recognizers on individual utterances. These maps indicate time-frequency points of the utterance that are most important for correct recognition of a target word. Our evaluation technique is not only suitable for standard classification tasks, but is also appropriate for structured prediction tasks like sequence-to-sequence models. Additionally, we use this approach to perform a large scale comparison of the importance maps created by our previously introduced technique using "bubble noise" to identify important points through correlation with a baseline approach based on smoothed speech energy and forced alignment. Our results show that the bubble analysis approach is better at identifying important speech regions than this baseline on 100 sentences from the AMI corpus.