eduzhai > Applied Sciences > Engineering >

Self-Supervised Representations Improve End-to-End Speech Translation

  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: End-to-end speech-to-text translation can provide a simpler and smallersystem but is facing the challenge of data scarcity. Pre-training methods canleverage unlabeled data and have been shown to be effective on data-scarcesettings. In this work, we explore whether self-supervised pre-trained speechrepresentations can benefit the speech translation task in both high- andlow-resource settings, whether they can transfer well to other languages, andwhether they can be effectively combined with other common methods that helpimprove low-resource end-to-end speech translation such as using a pre-trainedhigh-resource speech recognition system. We demonstrate that self-supervisedpre-trained features can consistently improve the translation performance, andcross-lingual transfer allows to extend to a variety of languages without orwith little tuning.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×