Language-Aware Multilingual Machine Translation with Self-Supervised Learning

TL;DR

This paper introduces a novel self-supervised learning approach with intra-distillation for multilingual machine translation, significantly enhancing performance by learning language-specific parameters and co-training with denoising tasks.

Contribution

It proposes a simple yet effective SSL task called concurrent denoising combined with intra-distillation to improve multilingual translation models.

Findings

Outperforms three state-of-the-art SSL methods by large margins

Achieves 11.3% and 3.7% improvements on 8-language and 15-language benchmarks

Highlights the importance of language-specific parameters in MMT

Abstract

Multilingual machine translation (MMT) benefits from cross-lingual transfer but is a challenging multitask optimization problem. This is partly because there is no clear framework to systematically learn language-specific parameters. Self-supervised learning (SSL) approaches that leverage large quantities of monolingual data (where parallel data is unavailable) have shown promise by improving translation performance as complementary tasks to the MMT task. However, jointly optimizing SSL and MMT tasks is even more challenging. In this work, we first investigate how to utilize intra-distillation to learn more *language-specific* parameters and then show the importance of these language-specific parameters. Next, we propose a novel but simple SSL task, concurrent denoising, that co-trains with the MMT task by concurrently denoising monolingual data on both the encoder and decoder. Finally, we apply intra-distillation to this co-training approach. Combining these two approaches significantly improves MMT performance, outperforming three state-of-the-art SSL methods by a large margin, e.g., 11.3\% and 3.7\% improvement on an 8-language and a 15-language benchmark compared with MASS, respectively