摘要
Chinese Phonetic-Character Conversion(CPCC) is an important issue in Chinese speech recognition and Chinese sentence keyboard input system. The approaches based on large corpus statistic Markov language model (such as bigram, trigram) become more and more popular today. This paper presents an improved Chinese word bigram, space-compressed Chinese word bigram, which stores the bi-word co-articulation frequency in the form of the bi-character co-articulation frequency. The bi-word co-articulation frequency is estimated from the bi-character co-articulation frequency library. The CPCC experiment with the improved Chinese word bigram shows: it can reach a higher correct conversion ratio with less space occupation.
Chinese Phonetic-Character Conversion(CPCC) is an important issue in Chinese speech recognition and Chinese sentence keyboard input system. The approaches based on large corpus statistic Markov language model (such as bigram, trigram) become more and more popular today. This paper presents an improved Chinese word bigram, space-compressed Chinese word bigram, which stores the bi-word co-articulation frequency in the form of the bi-character co-articulation frequency. The bi-word co-articulation frequency is estimated from the bi-character co-articulation frequency library. The CPCC experiment with the improved Chinese word bigram shows: it can reach a higher correct conversion ratio with less space occupation.
