Moses-based Chinese-Uyghur statistical machine translation systems

This paper is an initial explore to Chinese-Uyghur statistical machine translation. We experimented on three systems, which are based on moses framework, on Uyghur word-level corpus and morpheme-level corpus, then combine their 1-best translation hypothesis by constructing consensus network using TER alignment. The experiments show the combination of hypothesis that are from different systems can obtain a better result.