Syntactic Features for Evaluation of Machine Translation

Abstract

Automatic evaluation of machine transla-tion, based on computing n-gram similar-ity between system output and human ref-erence translations, has revolutionized the development of MT systems. We explore the use of syntactic information, includ-ing constituent labels and head-modier dependencies, in computing similarity be-tween output and reference. Our results show that adding syntactic information to the evaluation metric improves both sentence-level and corpus-level correla-tion with human judgments. 1