Improving example-based machine translation through morphological generalization and adaptation.

Abstract

Example Based Machine Translation (EBMT) is limited by the quantity and scope of its training data. Even with a reasonably large corpus, we will not have examples that cover everything we want to translate. This problem is especially severe in Arabic due to its rich morphology. We demonstrate a novel method that exploits the regular nature of Arabic morphology to increase the quality and coverage of machine translation. Through the use of generalization and rewrite rules, we are able to recover the English translation of phrases that do not exist in the training corpora. Furthermore, this system shows improvement in BLEU even with a training corpus of 1.4 million sentence pairs.