Efficient Speech Translation with Dynamic Latent Perceivers

TL;DR

This paper introduces a novel speech translation model using a Perceiver encoder with Dynamic Latent Access, reducing complexity and enabling flexible deployment without sacrificing translation quality.

Contribution

It proposes a Perceiver-based architecture with DLA training for efficient, scalable speech translation that matches Transformer performance across multiple language pairs.

Findings

Perceiver with DLA matches Transformer accuracy in MuST-C.

DLA enables flexible deployment across different computational budgets.

The approach reduces quadratic complexity of traditional Transformers.

Abstract

Transformers have been the dominant architecture for Speech Translation in recent years, achieving significant improvements in translation quality. Since speech signals are longer than their textual counterparts, and due to the quadratic complexity of the Transformer, a down-sampling step is essential for its adoption in Speech Translation. Instead, in this research, we propose to ease the complexity by using a Perceiver encoder to map the speech inputs to a fixed-length latent representation. Furthermore, we introduce a novel way of training Perceivers, with Dynamic Latent Access (DLA), unlocking larger latent spaces without any additional computational overhead. Speech-to-Text Perceivers with DLA can match the performance of Transformer baselines across three language pairs in MuST-C. Finally, a DLA-trained model is easily adaptable to DLA at inference, and can be flexibly deployed with various computational budgets, without significant drops in translation quality.