Semantically-informed syntactic machine translation: A tree-grafting approach

We describe a unified and coherent syntac-tic framework for supporting a semantically-informed syntactic approach to statistical ma-chine translation. Semantically enriched syn-tactic tags assigned to the target-language training texts improved translation quality. The resulting system significantly outper-formed a linguistically naive baseline model (Hiero), and reached the highest scores yet re-ported on the NIST 2009 Urdu-English trans-lation task. This finding supports the hy-pothesis (posed by many researchers in the MT community, e.g., in DARPA GALE) that both syntactic and semantic information are critical for improving translation quality—and further demonstrates that large gains can be achieved for low-resource languages with dif-ferent word order than English.