Leveraging Unit Language Guidance to Advance Speech Modeling in Textless Speech-to-Speech Translation

TL;DR

This paper introduces a unit language approach for textless speech-to-speech translation, addressing cross-modal and cross-lingual challenges, and demonstrates significant improvements in multilingual speech translation performance.

Contribution

The paper proposes a novel unit language representation and task prompt modeling to improve speech modeling in textless S2ST, overcoming key cross-modal and cross-lingual challenges.

Findings

Significant performance improvements over baseline models.

Achieves results comparable to text-based models.

Effective mitigation of source-target unit language conflict.

Abstract

The success of building textless speech-to-speech translation (S2ST) models has attracted much attention. However, S2ST still faces two main challenges: 1) extracting linguistic features for various speech signals, called cross-modal (CM), and 2) learning alignment of difference languages in long sequences, called cross-lingual (CL). We propose the unit language to overcome the two modeling challenges. The unit language can be considered a text-like representation format, constructed using $n$-gram language modeling. We implement multi-task learning to utilize the unit language in guiding the speech modeling process. Our initial results reveal a conflict when applying source and target unit languages simultaneously. We propose task prompt modeling to mitigate this conflict. We conduct experiments on four languages of the Voxpupil dataset. Our method demonstrates significant improvements over a strong baseline and achieves performance comparable to models trained with text.