Explaining the Attention Mechanism of End-to-End Speech Recognition Using Decision Trees
This work addresses the interpretability of attention mechanisms for researchers in speech recognition, but it is incremental as it provides insights without introducing a new method or achieving performance gains.
The study tackled the problem of understanding the attention mechanism in end-to-end speech recognition by using decision trees to analyze its behavior, finding that attention levels are primarily influenced by previous states rather than encoder or decoder patterns, and that the default mechanism focuses on closer states but struggles with long-term dependencies.
The attention mechanism has largely improved the performance of end-to-end speech recognition systems. However, the underlying behaviours of attention is not yet clearer. In this study, we use decision trees to explain how the attention mechanism impact itself in speech recognition. The results indicate that attention levels are largely impacted by their previous states rather than the encoder and decoder patterns. Additionally, the default attention mechanism seems to put more weights on closer states, but behaves poorly on modelling long-term dependencies of attention states.