An application of statistical optimization with dynamic programming to phonemic-input-to-character conversion for Chinese.

This paper describes a method for the conversion of Chinese text written in phonemic transliteration into Chinese character text. The method uses character bigram probabilities estimated from a large corpus of Chinese text to compute the most likely path through the lattice of possible character transliterations. Dynamic programming is used to prune the search. Thus the method is statistical, like previous work by Lin and Tsai (1987), but it is algorithmically simpler than their method. Performance scores are given which demonstrate that the proposed method produces good results.