DiDi's Machine Translation System for WMT2020

TL;DR

This paper details DiDi AI Labs' Chinese-English translation system for WMT2020, utilizing Transformer models and various enhancement techniques to achieve a BLEU score of 36.6.

Contribution

The paper introduces a comprehensive system combining multiple techniques like data filtering, back-translation, and ensembling for improved machine translation performance.

Findings

Achieved BLEU score of 36.6 on Chinese-English translation

Demonstrated effectiveness of combined enhancement techniques

Showcased state-of-the-art results in WMT2020 translation task

Abstract

This paper describes DiDi AI Labs' submission to the WMT2020 news translation shared task. We participate in the translation direction of Chinese->English. In this direction, we use the Transformer as our baseline model, and integrate several techniques for model enhancement, including data filtering, data selection, back-translation, fine-tuning, model ensembling, and re-ranking. As a result, our submission achieves a BLEU score of $36.6$ in Chinese->English.