Multilingual Neural Machine Translation with Knowledge Distillation

TL;DR

This paper introduces a knowledge distillation method to improve multilingual neural machine translation, enabling a single model to outperform or match multiple individual models across many languages.

Contribution

It proposes a distillation-based training approach where a multilingual model learns from individual language pair models, enhancing accuracy and efficiency.

Findings

Multilingual model handles up to 44 languages effectively.

Achieves comparable or better accuracy than individual models.

Demonstrates effectiveness on multiple translation datasets.

Abstract

Multilingual machine translation, which translates multiple languages with a single model, has attracted much attention due to its efficiency of offline training and online serving. However, traditional multilingual translation usually yields inferior accuracy compared with the counterpart using individual models for each language pair, due to language diversity and model capacity limitations. In this paper, we propose a distillation-based approach to boost the accuracy of multilingual machine translation. Specifically, individual models are first trained and regarded as teachers, and then the multilingual model is trained to fit the training data and match the outputs of individual models simultaneously through knowledge distillation. Experiments on IWSLT, WMT and Ted talk translation datasets demonstrate the effectiveness of our method. Particularly, we show that one model is enough to handle multiple languages (up to 44 languages in our experiment), with comparable or even better accuracy than individual models.