基于卷积神经网络的英语点餐机器人智能交互研究  

Research on Intelligent Interaction of English Order Robot Based on Convolutional Neural Network

作　　者：魏文萍[1] WEI Wenping(Xianyang Vocational and Technical College,Xianyang,Shaanxi 712000,China)

机构地区：[1]咸阳职业技术学院,陕西咸阳712000

出　　处：《自动化与仪器仪表》2022年第7期252-256,共5页Automation & Instrumentation

基　　金：陕西省教育科学"十四五"规划2021年度课题《关中方言及文化在留学生教学与管理中的渗透与传播研究》(SGH21Y0587);河南省高校人文社会科学研究项目《人工智能机器翻译背景下的英语翻译人才培养方式创新研究》(2022-ZDJH-00494)。

摘　　要：针对传统英语点餐机器人情感交互能力弱,智能化水平低的问题,构建BiLSTM-SeqGAN网络的英语点餐交互系统。通过Seq2Seq网络和BiLSTM网络分别对输入英语语句进行编码和双向信息特征提取;之后在此基础上加入约束型SeqGAN网络架构进行语义信息解码,并提出主旨型注意力模式提取解码后的语义信息;对生成装置参数进行调整后,即可生成英语情感交互回复。仿真结果表明,本算法的困惑指标仅为0.25,比另外两种算法更低,而本算法精准度高达92.6%,说明该算法在点餐机器人中可生成自然、真实的情感交互回复,交互能力强,智能化水平显著提高。In view of the weak emotional interaction ability and low intelligence level of the traditional English ordering robot,the English ordering interaction system of BiLSTM-SeqGAN network is built.The input English statements are encoded and bidirectional information features are extracted through Seq2 Seq network and BiLSTM network.Later,the constrained SeqGAN network architecture is added for semantic information decoding,and the thematic attention mode is proposed.After adjusting the generated device parameters,the English emotional interaction reply can be generated.Simulation results show that the confusion index of this algorithm is only 0.25,which is lower than the other two algorithms,while the accuracy of this algorithm is as high as 92.6%,indicating that the algorithm can generate natural and real emotional interaction response in the ordering robot,strong interaction ability and intelligence level is significantly improved.

关 键 词：Seq2Seq BiLSTM网络 点餐机器人 情感交互 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]