TransQuest at WMT2020: Sentence-Level Direct Assessment

TL;DR

This paper introduces a transformer-based quality estimation framework for sentence-level translation assessment, achieving state-of-the-art results and winning all language pair categories in WMT 2020.

Contribution

The paper presents a simple yet effective QE framework based on cross-lingual transformers, with ensemble and data augmentation techniques, outperforming previous baselines.

Findings

Achieved state-of-the-art results in WMT 2020 shared task

Outperformed the baseline OpenKiwi in all language pairs

Winning solution across all evaluated language pairs

Abstract

This paper presents the team TransQuest's participation in Sentence-Level Direct Assessment shared task in WMT 2020. We introduce a simple QE framework based on cross-lingual transformers, and we use it to implement and evaluate two different neural architectures. The proposed methods achieve state-of-the-art results surpassing the results obtained by OpenKiwi, the baseline used in the shared task. We further fine tune the QE framework by performing ensemble and data augmentation. Our approach is the winning solution in all of the language pairs according to the WMT 2020 official results.