Language Model-Driven Unsupervised Neural Machine Translation

TL;DR

This paper introduces a novel unsupervised neural machine translation approach that leverages language models and a weighted data scoring mechanism to improve translation quality, outperforming previous systems on standard benchmarks.

Contribution

It proposes a new method combining language models with weighted synthetic data scoring for unsupervised NMT, enhancing translation accuracy.

Findings

Outperforms prior systems by over 3 BLEU points on WMT benchmarks.

Effectively utilizes language models to reduce noise in synthetic data.

Demonstrates significant improvements in English-French, German, and Russian translation tasks.

Abstract

Unsupervised neural machine translation(NMT) is associated with noise and errors in synthetic data when executing vanilla back-translations. Here, we explicitly exploits language model(LM) to drive construction of an unsupervised NMT system. This features two steps. First, we initialize NMT models using synthetic data generated via temporary statistical machine translation(SMT). Second, unlike vanilla back-translation, we formulate a weight function, that scores synthetic data at each step of subsequent iterative training; this allows unsupervised training to an improved outcome. We present the detailed mathematical construction of our method. Experimental WMT2014 English-French, and WMT2016 English-German and English-Russian translation tasks revealed that our method outperforms the best prior systems by more than 3 BLEU points.