Relative Positional Encoding for Speech Recognition and Direct Translation

TL;DR

This paper introduces a relative positional encoding scheme for Speech Transformers, improving speech recognition and translation accuracy by better modeling acoustic input variability and data utilization.

Contribution

The paper adapts relative position encoding to Speech Transformers, enhancing their ability to handle speech data's variable distributions and segmentation quality.

Findings

Achieved best recognition on Switchboard benchmark

Set new state-of-the-art in MuST-C speech translation

Better utilization of synthetic data and segmentation variability

Abstract

Transformer models are powerful sequence-to-sequence architectures that are capable of directly mapping speech inputs to transcriptions or translations. However, the mechanism for modeling positions in this model was tailored for text modeling, and thus is less ideal for acoustic inputs. In this work, we adapt the relative position encoding scheme to the Speech Transformer, where the key addition is relative distance between input states in the self-attention network. As a result, the network can better adapt to the variable distributions present in speech data. Our experiments show that our resulting model achieves the best recognition result on the Switchboard benchmark in the non-augmentation condition, and the best published result in the MuST-C speech translation benchmark. We also show that this model is able to better utilize synthetic data than the Transformer, and adapts better to variable sentence segmentation quality for speech translation.