Sequence-to-Sequence Models Can Directly Translate Foreign Speech

TL;DR

This paper demonstrates that a modified sequence-to-sequence neural network can directly translate speech from one language to text in another without intermediate transcription, achieving state-of-the-art results on Spanish-English translation.

Contribution

It introduces a unified end-to-end speech translation model that outperforms traditional cascaded systems and leverages multi-task training for improved accuracy.

Findings

State-of-the-art BLEU score on Fisher Spanish-English dataset

End-to-end model outperforms cascade systems by 1.8 BLEU points

Multi-task training with shared encoder improves performance by 1.4 BLEU points

Abstract

We present a recurrent encoder-decoder deep neural network architecture that directly translates speech in one language into text in another. The model does not explicitly transcribe the speech into text in the source language, nor does it require supervision from the ground truth source language transcription during training. We apply a slightly modified sequence-to-sequence with attention architecture that has previously been used for speech recognition and show that it can be repurposed for this more complex task, illustrating the power of attention-based models. A single model trained end-to-end obtains state-of-the-art performance on the Fisher Callhome Spanish-English speech translation task, outperforming a cascade of independently trained sequence-to-sequence speech recognition and machine translation models by 1.8 BLEU points on the Fisher test set. In addition, we find that making use of the training data in both languages by multi-task training sequence-to-sequence speech translation and recognition models with a shared encoder network can improve performance by a further 1.4 BLEU points.