基于NLP的语音敏感词智能识别方法

Intelligent recognition method of speech sensitive words based on NLP

针对语言处理技术精度较差,语音敏感词识别率较低的问题,提出基于自然语言处理技术的语音敏感词智能识别方法。以自然语言处理技术为基础,构建语音处理模型;通过语音分析提取语音中包含的特征参数,依靠决策树方法,设计敏感词识别算法;通过语音敏感词的敏感度计算,将敏感词使用特殊字符标注,实现语音智能的最终识别。实验结果表明:文中设计的智能识别方法,明显优于GSV-SVM和i-vector+PLDA两种识别方法,在相同测试环境中,敏感词识别率分别提升了4.84%、15.24%。

For the problem of poor accuracy of language processing technology and low recognition rate of sensitive speech words,an intelligent recognition method of speech sensitive words based on natural language processing technology is proposed.The speech processing model is constructed based on the natural language processing technology,and the characteristic parameters are extracted based on the speech analysis.The sensitive word recognition algorithm is designed on the basis of decision tree to label the sensitive words with special characters,all of which could realize the final recognition of speech intelligence.The experiment results show that the intelligent recognition method designed in the paper is significantly better than GSV-SVM and i-vector+PLDA.Moreover,the recognition rate of sensitive words is increased by 4.84%and 15.24%in the same test environment.