Understanding Self-Attention of Self-Supervised Audio Transformers
Computation and Language
2020-08-12 v2
Abstract
Self-supervised Audio Transformers (SAT) enable great success in many downstream speech applications like ASR, but how they work has not been widely explored yet. In this work, we present multiple strategies for the analysis of attention mechanisms in SAT. We categorize attentions into explainable categories, where we discover each category possesses its own unique functionality. We provide a visualization tool for understanding multi-head self-attention, importance ranking strategies for identifying critical attention, and attention refinement techniques to improve model performance.
Cite
@article{arxiv.2006.03265,
title = {Understanding Self-Attention of Self-Supervised Audio Transformers},
author = {Shu-wen Yang and Andy T. Liu and Hung-yi Lee},
journal= {arXiv preprint arXiv:2006.03265},
year = {2020}
}
Comments
Accepted by INTERSPEECH 2020, ICML 2020 Workshop on Self-supervision in Audio and Speech