Bridging the Data Gap between Training and Inference for Unsupervised Neural Machine Translation

TL;DR

This paper addresses the source data discrepancy in unsupervised neural machine translation caused by style and content gaps, proposing an online self-training method that improves translation quality across multiple language pairs.

Contribution

It introduces an online self-training approach that reduces style and content gaps in UNMT, leading to improved translation performance over existing methods.

Findings

Outperforms strong baselines like XLM and MASS

Effectively narrows style and content gaps in data

Improves translation accuracy across multiple language pairs

Abstract

Back-translation is a critical component of Unsupervised Neural Machine Translation (UNMT), which generates pseudo parallel data from target monolingual data. A UNMT model is trained on the pseudo parallel data with translated source, and translates natural source sentences in inference. The source discrepancy between training and inference hinders the translation performance of UNMT models. By carefully designing experiments, we identify two representative characteristics of the data gap in source: (1) style gap (i.e., translated vs. natural text style) that leads to poor generalization capability; (2) content gap that induces the model to produce hallucination content biased towards the target language. To narrow the data gap, we propose an online self-training approach, which simultaneously uses the pseudo parallel data {natural source, translated target} to mimic the inference scenario. Experimental results on several widely-used language pairs show that our approach outperforms two strong baselines (XLM and MASS) by remedying the style and content gaps.