XLM-T: Scaling up Multilingual Machine Translation with Pretrained Cross-lingual Transformer Encoders
This work provides an incremental improvement for researchers and practitioners working on multilingual machine translation by demonstrating the effectiveness of pre-trained cross-lingual encoders.
This paper introduces XLM-T, a multilingual machine translation system that initializes its Transformer encoder with an off-the-shelf pretrained cross-lingual model. This approach significantly improves performance on WMT with 10 language pairs and OPUS-100 with 94 pairs, even outperforming strong baselines that use back-translation.
Multilingual machine translation enables a single model to translate between different languages. Most existing multilingual machine translation systems adopt a randomly initialized Transformer backbone. In this work, inspired by the recent success of language model pre-training, we present XLM-T, which initializes the model with an off-the-shelf pretrained cross-lingual Transformer encoder and fine-tunes it with multilingual parallel data. This simple method achieves significant improvements on a WMT dataset with 10 language pairs and the OPUS-100 corpus with 94 pairs. Surprisingly, the method is also effective even upon the strong baseline with back-translation. Moreover, extensive analysis of XLM-T on unsupervised syntactic parsing, word alignment, and multilingual classification explains its effectiveness for machine translation. The code will be at https://aka.ms/xlm-t.