A Maximum Entropy Approach to Chinese Pin Yin-To-Character Conversion

This paper introduces a new approach based upon maximum entropy (ME) frame to solve the Pinyin-to-character (PTC) conversation problem. Mostly there is more than one Chinese characters share the same Pinyin. The task of PTC algorithm is to distinguish such kind ambiguity. PTC can be regards as to classify a Pinyin to a special character according the context which is represented as feature in ME. By taking the advantage of ME, the local and non-local information are included, so the conversation performance is improved. Experiments show that 87% hit rate (without tone) is achieved.