Exploring Spanish-morphology effects on Chinese–Spanish SMT

This paper presents some statistical machine translation results among English, Spanish and Chinese, and focuses on exploring Spanish-morphology effects on the Chinese to Spanish translation task. Although not strictly comparable, it is observed that by reducing Spanish morphology the accuracy achieved in the Chinese to Spanish translation task becomes comparable to the one achieved in the Chinese to English task. Further experimentation on approaching the problem of generating Spanish morphology as a translation task by itself is also performed, and results discussed. All experiments have been carried out by using a trilingual parallel corpus extracted from the Bible.