Chinese-to-English Backward Machine Transliteration

It is challenging to transliterate named entities across languages. It is even more challenging to backward transliterate the transliterated term into its original form. This paper addresses the problem of backward translating person name from Chinese to its English counterpart. We propose a statistical backward transliteration method. Our method uses English sub-syllable and Chinese syllable as the basic units for the Chinese-English transliteration model and the English language model. Experiments show that our method is promising.