Fine-grained linguistic evaluation for state-of-the-art Machine Translation

TL;DR

This paper presents a detailed linguistic test suite for evaluating German-English machine translation systems, revealing strengths and weaknesses across various phenomena and comparing system performances.

Contribution

It introduces a comprehensive linguistic evaluation framework with detailed analysis for state-of-the-art MT systems, highlighting specific linguistic challenges and system improvements.

Findings

Tohoku and Huoshan systems outperform others in test suite accuracy.

All systems struggle with idioms, resultative predicates, and pluperfect.

Most WMT19 systems improved in the latest evaluation.

Abstract

This paper describes a test suite submission providing detailed statistics of linguistic performance for the state-of-the-art German-English systems of the Fifth Conference of Machine Translation (WMT20). The analysis covers 107 phenomena organized in 14 categories based on about 5,500 test items, including a manual annotation effort of 45 person hours. Two systems (Tohoku and Huoshan) appear to have significantly better test suite accuracy than the others, although the best system of WMT20 is not significantly better than the one from WMT19 in a macro-average. Additionally, we identify some linguistic phenomena where all systems suffer (such as idioms, resultative predicates and pluperfect), but we are also able to identify particular weaknesses for individual systems (such as quotation marks, lexical ambiguity and sluicing). Most of the systems of WMT19 which submitted new versions this year show improvements.