Extending Multilingual Machine Translation through Imitation Learning

TL;DR

This paper introduces Imit-MNMT, a novel imitation learning approach to extend multilingual neural machine translation models to new languages using pseudo-parallel data, effectively improving translation quality and reducing forgetting.

Contribution

The paper presents a new imitation learning framework for multilingual translation that leverages expert-generated pseudo-data to incorporate new languages without degrading existing performance.

Findings

Significant improvement in translation quality for new languages.

Reduced catastrophic forgetting in multilingual models.

Effective use of imitation learning in NLP tasks.

Abstract

Despite the growing variety of languages supported by existing multilingual neural machine translation (MNMT) models, most of the world's languages are still being left behind. We aim to extend large-scale MNMT models to incorporate a new language, enabling translations between this new language and all previously supported languages, even in the challenging scenario where only a parallel corpus between the new language and English is available. Previous methods, such as continued training on parallel data including the new language, often suffer from catastrophic forgetting, which degrades performance on other languages. We propose a novel approach Imit-MNMT which treats this task as an imitation learning problem, a technique widely used in computer vision but less explored in natural language processing. Specifically, we leverage an expert model to generate pseudo-parallel corpora between the new language and the existing languages. We then introduce a data distribution imitation strategy using language-specific weighting, alongside a translation behavior imitation mechanism. Extensive experiments show that our approach significantly improves translation performance between the new and existing languages while mitigating catastrophic forgetting.