Can We Achieve High-quality Direct Speech-to-Speech Translation without Parallel Speech Data?

TL;DR

This paper introduces ComSpeech, a novel model that integrates pretrained S2TT and TTS models for direct speech-to-speech translation, and proposes ComSpeech-ZS for zero-shot translation without parallel speech data, achieving competitive results.

Contribution

The paper presents a new composite S2ST model and a zero-shot training method that leverages existing pretrained models and data, reducing reliance on parallel speech data.

Findings

ComSpeech outperforms previous models in translation quality and speed when parallel data is available.

ComSpeech-ZS achieves near state-of-the-art results without parallel speech data.

The methods effectively utilize pretrained models and contrastive learning for zero-shot translation.

Abstract

Recently proposed two-pass direct speech-to-speech translation (S2ST) models decompose the task into speech-to-text translation (S2TT) and text-to-speech (TTS) within an end-to-end model, yielding promising results. However, the training of these models still relies on parallel speech data, which is extremely challenging to collect. In contrast, S2TT and TTS have accumulated a large amount of data and pretrained models, which have not been fully utilized in the development of S2ST models. Inspired by this, in this paper, we first introduce a composite S2ST model named ComSpeech, which can seamlessly integrate any pretrained S2TT and TTS models into a direct S2ST model. Furthermore, to eliminate the reliance on parallel speech data, we propose a novel training method ComSpeech-ZS that solely utilizes S2TT and TTS data. It aligns representations in the latent space through contrastive learning, enabling the speech synthesis capability learned from the TTS data to generalize to S2ST in a zero-shot manner. Experimental results on the CVSS dataset show that when the parallel speech data is available, ComSpeech surpasses previous two-pass models like UnitY and Translatotron 2 in both translation quality and decoding speed. When there is no parallel speech data, ComSpeech-ZS lags behind \name by only 0.7 ASR-BLEU and outperforms the cascaded models.