Chinese-uyghur statistical machine translation: The initial explorations

In this paper, we present results of initial explorations to a phrase-based statistical machine translation system for a new language pair, namely Chinese-Uyghur. They are very different from each other, the characters of the former almost are hieroglyphics, morpheme processing don't work at all, but the latter is an agglutinative language with very productive inflectional and derivational word-formation processes. To make them more similar, we reorder Chinese sentence structures from SVO to SOV and split Uyghur words into morphemes. The experiments show reordering Chinese sentence structure and properly splitting granularity for Uyghur can effectively improve the performances of translation system.