Scoring the Translation: On Target Automatic Keyword-Based Evaluation of Machine Translation in the Sports Domain

Abstract

We take a closer look at the results of a recent translation shared task at WMT 2025 (the Conference on Machine Translation) and analyse the errors in the output of the four highest-scoring systems. We revise the automatic evaluation method used in Sigurðsson et al. (2025) and compare it to manual evaluation of six machine translation systems. We find that our results are in line with the manual evaluation, indicating that the test suite can be well suited for evaluating machine translation in this domain. Finally, we publish a list of domain-specific sports terms, namely, in the domains of basketball, chess, football, golf and gymnastics.

Resources

Details