基于智能终端的语音阅读交互研究

Research on speech reading interaction based on intelligent Terminal

为提高语音阅读交互系统中回复内容的匹配速率和准确率,从信息匹配的算法入手,提出基于attention机制对传统seq2seq模型进行改进,并引入beam search算法对语音识别结果进行剪枝,选择出最佳的匹配内容输出;接着进行语音阅读交互系统设计,选择语音识别系统、设定文本信息向编码信息转换的方式、对基于改进seq2seq模型的答复匹配设计;最后为验证改进后的seq2seq模型在语音阅读交互系统中的优势,配置语音阅读交互系统运行的实验环境,并进行算法验证实验。最终的结果证明,改进后的算法损失函数的收敛性性能更好;交互时回复信息的相关性明显高于改进前,且回复的信息的内容更为丰富。

In order to improve the matching rate and accuracy of the response content in the speech reading interactive system,this paper proposes an improved seq2seq model based on attention mechanism,and introduces beam search algorithm to prune the speech recognition results,then,the design of speech recognition system,the way of converting text information to coding information,and the response matching design based on improved seq2seq model are carried out Finally,in order to verify the advantages of the improved seq2seq model in the speech reading interactive system,the experimental environment of the speech reading interactive system is configured,and the Algorithm verification experiment is carried out.The results show that the improved algorithm has better convergence performance,and the correlation of returned information is obviously higher than that before the improvement,and the content of returned information is more abundant.