InfoXLM: An Information-Theoretic Framework for Cross-Lingual Language Model Pre-Training

TL;DR

This paper introduces an information-theoretic framework for cross-lingual language model pre-training, utilizing mutual information maximization and contrastive learning to enhance multilingual transferability and performance.

Contribution

It proposes a novel pre-training framework based on mutual information and contrastive learning, improving cross-lingual representations over existing methods.

Findings

Achieves better performance on multiple benchmarks

Effectively leverages monolingual and parallel corpora

Enhances cross-lingual transferability

Abstract

In this work, we present an information-theoretic framework that formulates cross-lingual language model pre-training as maximizing mutual information between multilingual-multi-granularity texts. The unified view helps us to better understand the existing methods for learning cross-lingual representations. More importantly, inspired by the framework, we propose a new pre-training task based on contrastive learning. Specifically, we regard a bilingual sentence pair as two views of the same meaning and encourage their encoded representations to be more similar than the negative examples. By leveraging both monolingual and parallel corpora, we jointly train the pretext tasks to improve the cross-lingual transferability of pre-trained models. Experimental results on several benchmarks show that our approach achieves considerably better performance. The code and pre-trained models are available at https://aka.ms/infoxlm.