XLM-T: Scaling up Multilingual Machine Translation with Pretrained Cross-lingual Transformer Encoders

TL;DR

XLM-T leverages pretrained cross-lingual Transformer encoders for multilingual machine translation, significantly improving performance across multiple datasets and demonstrating versatility in related multilingual tasks.

Contribution

The paper introduces XLM-T, a novel approach that fine-tunes pretrained cross-lingual models for multilingual translation, outperforming traditional randomly initialized models.

Findings

Significant BLEU score improvements on WMT and OPUS-100 datasets.

Effective even with strong back-translation baselines.

Analysis shows benefits in syntactic parsing, word alignment, and classification.

Abstract

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.