摘要
为使英语聊天机器人的聊天内容更加生动有趣,将结合循环神经网络的特性,构建基于Seq2Seq的英语聊天机器人模型。首先,具体分析长期记忆神经网络LSTM的基本原理和网络结构,然后分别设计一个LSTM-LDA主题模型和深度语言模型,将LSTM-LDA的主题信息和语言模型中的语义信息融合后,再进行编码处理,最终得到一个完整的聊天机器人模型。实验结果表明,LSTM-LDA模型获取主题信息的准确率高达96.5%,远远高于其他模型。且通过模型训练和生成结果发现,设计的聊天机器人可根据主题信息进行聊天,聊天内容趣味性更强。
In order to make the chat content of the English chatbot more vivid and interesting,a Seq2 Seq-based English chatbot model will be built,combined with the characteristics of the circular neural network.First,the basic principles and network structure of long-term memory neural network are analyzed,and then a LSTM-LDA theme model and deep language model respectively,integrate the theme information of LSTM-LDA and semantic information in the language model,and then encoding processing,and finally get a complete chatbot model.The experimental results show that the LSTM-LDA model obtains subject information with 96.5% accuracy,much higher than the other models.Moreover,through model training and generating results,it is found that the designed chatbot can chat according to the theme information,and the chat content is more interesting.
作者
陈潇艺
CHEN Xiaoyi(Sichuan Vocational College of Science and Technology,Meishan,Sichuan 620500,China)
出处
《自动化与仪器仪表》
2022年第7期242-246,251,共6页
Automation & Instrumentation
基金
四川省教育项目《“应用型人才”视阈下高职《学前儿童英语教育》课程教学问题及对策研究——以四川地级市为例》(SCGJ2021-25)。
