人工智能技术在自然语音纠错与反馈系统设计中的应用

Application of Artificial Intelligence Technology in the Design of Natural Speech Error Correction and Feedback System

为实现自然语音纠错,提升自然语音识别与拼读的正确率,研究人工智能技术在自然语音纠错与反馈系统设计中的应用。设计由前端学习单元与后端支撑单元组成的自然语音纠错与反馈系统,预处理采集到的自然语音片段,基于片段间距离划分因素,提取自然语音片段特征,采用隐马尔可夫模型识别自然语音,基于B2规范语料,采用动态时间归整方法纠错与评分识别到的自然语音,通过反馈模块将识别、纠错、评分结果反馈给用户。对比实验的结果表明,设计的自然语音纠错与反馈系统的语音识别率高于95%,纠错结果与实际错误一致,可提升自然语音拼读的正确率。

In order to realize natural speech error correction and improve the accuracy of natural speech recognition and spelling,the application of artificial intelligence technology in the design of natural speech error correction and feedback system is studied.A natural speech error correction and feedback system composed of front-end learning unit and back-end support unit is designed.The collected natural speech segments are preprocessed.Based on the distance between segments,the features of natural speech segments are extracted.The hidden Markov model is used to recognize the natural speech.Based on the B2 standard corpus,the dynamic time integration method is used to correct and grade the natural speech.Through the feedback module,the recognition,error correction and scoring results are fed back to users.The experimental results show that the speech recognition rate of the designed natural speech error correction and feedback system is higher than 95%,and the error correction result is consistent with the actual error,which can improve the accuracy of natural speech spelling.