Cascaded Models With Cyclic Feedback For Direct Speech Translation

TL;DR

This paper introduces a cyclic feedback technique for cascaded speech translation models that leverages in-domain data and improves performance by self-training and fine-tuning, achieving significant BLEU score gains.

Contribution

It proposes a novel cyclic feedback approach for cascaded speech translation models that enhances performance using in-domain data and self-training strategies.

Findings

Up to 3.8 BLEU point improvements on LibriVoxDeEn

Up to 5.1 BLEU point improvements on CoVoST

Enhanced tolerance to spelling variations in MT component

Abstract

Direct speech translation describes a scenario where only speech inputs and corresponding translations are available. Such data are notoriously limited. We present a technique that allows cascades of automatic speech recognition (ASR) and machine translation (MT) to exploit in-domain direct speech translation data in addition to out-of-domain MT and ASR data. After pre-training MT and ASR, we use a feedback cycle where the downstream performance of the MT system is used as a signal to improve the ASR system by self-training, and the MT component is fine-tuned on multiple ASR outputs, making it more tolerant towards spelling variations. A comparison to end-to-end speech translation using components of identical architecture and the same data shows gains of up to 3.8 BLEU points on LibriVoxDeEn and up to 5.1 BLEU points on CoVoST for German-to-English speech translation.