On the Locality of Attention in Direct Speech Translation

TL;DR

This paper investigates the role of self-attention in direct speech translation, revealing local patterns and proposing a more efficient local attention mechanism that maintains performance while reducing computational complexity.

Contribution

It introduces a local attention mechanism based on layer-wise analysis, improving efficiency without sacrificing translation quality in speech tasks.

Findings

Local attention captures key token interactions effectively.

Proposed method matches baseline performance.

Efficiency improves by skipping unnecessary attention weights.

Abstract

Transformers have achieved state-of-the-art results across multiple NLP tasks. However, the self-attention mechanism complexity scales quadratically with the sequence length, creating an obstacle for tasks involving long sequences, like in the speech domain. In this paper, we discuss the usefulness of self-attention for Direct Speech Translation. First, we analyze the layer-wise token contributions in the self-attention of the encoder, unveiling local diagonal patterns. To prove that some attention weights are avoidable, we propose to substitute the standard self-attention with a local efficient one, setting the amount of context used based on the results of the analysis. With this approach, our model matches the baseline performance, and improves the efficiency by skipping the computation of those weights that standard attention discards.