基于全信息自然语言理解的语音识别后文本处理

将全信息(自然语言的语法、语义和语用信息)自然语言理解应用到可靠语音功能的研究中，提出在语音识别之后增加一个全信息自然语言理解的文本后处理功能，通过深入分析词语的语法信息（位置、识别稳定度）、语义信息（语句目标含义）和语用信息（语境和谐度）对语音识别结果语句进行评估、检错和纠错，最终输出优化语句。实验结果表明，增加全信息自然语言理解后处理可使识别正确率获得较好改善。

CI-NLU Based Text Processing After Speech Recognition

Applying the theory of CI-NLU (comprehensive information based natural language understanding, which protests that natural language could be understood through analyzing comprehensive information, the entirety of syntactic, semantic and pragmatic information.) into the research of reliable Speech Recognition, a new method of adding a text post-processing function of CI-NLU is put forward. The syntactic information of words (location and Speech Recognition stability), semantic information (meaning of sentence) and pragmatic information (harmony of word and its environment) are analyzed to check the sentence reliability of Speech Recognition result. Testing results have shown that the accuracy could be improved to a good extent.