Machine Translation Customization via Automatic Training Data Selection from the Web

TL;DR

This paper presents a method for customizing neural machine translation systems to specific domains by automatically selecting relevant training data from the Web using document classifiers, leading to improved performance with less data.

Contribution

The authors introduce a novel data selection approach using monolingual target data and document classifiers to enhance domain-specific neural machine translation.

Findings

Outperforms top systems on WMT-18 News translation benchmark

Uses less data and smaller models than state-of-the-art systems

Achieves better domain adaptation for machine translation

Abstract

Machine translation (MT) systems, especially when designed for an industrial setting, are trained with general parallel data derived from the Web. Thus, their style is typically driven by word/structure distribution coming from the average of many domains. In contrast, MT customers want translations to be specialized to their domain, for which they are typically able to provide text samples. We describe an approach for customizing MT systems on specific domains by selecting data similar to the target customer data to train neural translation models. We build document classifiers using monolingual target data, e.g., provided by the customers to select parallel training data from Web crawled data. Finally, we train MT models on our automatically selected data, obtaining a system specialized to the target domain. We tested our approach on the benchmark from WMT-18 Translation Task for News domains enabling comparisons with state-of-the-art MT systems. The results show that our models outperform the top systems while using less data and smaller models.