Towards Unsupervised Speech-to-Speech Translation

Anonymous

Keywords: speech translation, unsupervised, UASR, UMT, UTTS

Abstract: Most of the speech-to-speech translation models heavily rely on parallel data, which is hard to collect especially for low-resource languages. To tackle this issue, we propose to build a speech-to-speech translation system without leveraging any kind of paired data. To the best of our knowledge, this work is the first one that has successfully built a speech-to-speech translation system under an unsupervised scenario. We use fully unpaired data to train our unsupervised system and make comparable results with the other supervised methods proposed just a few years ago. Furthermore, to demonstrate that our method can generalize well across different languages, we evaluate our system on CVSS, a multi-lingual speech-to-speech corpus, and get promising results in different translation directions.

Paper Type: long

Research Area: Speech recognition, text-to-speech and spoken language understanding