汉字层面的汉语语言统计模型

A Statistic Language Model on Chinese at Character Level

利用汉语语言的统计规律对汉语连续语音识别系统的声学到语音学的结果进行纠错及音字转换具有重要意义.本文介绍一个采用统计方法实现的字层面的三元语言模型.它较为充分地利用了前端声学匹配的结果,对于通常的汉语短语及句子的声学识别结果,具有很高的纠错率及转换率.

It is very important and useful to improve the accuracy of speech recognition and understanding by taking use of the statistics of the language of interesting. A good language model can reduce the mistakes and perform the transformation from sequences of pronunciation to senquences of language characters. This article introduces a language model at Chinere charater level,which works well with general Chinese phrases and sentenees.