Research And Design Of Personalized Dialog System Based Service Robot Cloud Platform

With the development of artificial intelligence technology,service robots have gradually entered thousands of households.Human-computer interactive dialogue system,based on natural language,makes it convenient for human-computer interaction,and plays an important role in the popularization and application of service robots.However,there are still some problems in the existing robot dialogue system,such as the imperfect construction of user model,less reply information generated and the same reply content for different users.In order to improve the user experience of service robot dialog system,based on the review of the dialog generation technology,this paper researches on the construction of user model and personalized dialog generation model of the dialog system,designs and builds the personalized dialog system of service robot based on the cloud platform.The main contents and innovations of this paper are summarized as follows.Firstly,in the aspect of user model construction,this paper proposes to use the model of topics to build the user model.According to the characteristics of short text information in user voice interaction,Twitter-LDA model is used to mine user's interesting topics.At the same time,for the purpose of ensuring the accuracy of user model,the original topic model is improved,and the dynamic topic model is proposed considering the time factor of user interaction.Using the data from social media platforms and the consistency of communication among people in daily life,the topic model is trained with open data.Secondly,aiming at the problem that the traditional generative dialog system generates a single reply content and less information,this paper proposes a personalized dialog model that integrates user model.A special network structure is designed,which combines memory network with Transfomer based on full attention mechanism.Memory network is used to obtain context of dialogue historry and user model information.It is called Personalized Transformer Memory Networks.Throughthe analysis of experiments,compared with the dialogue model generated by traditional end-to-end neural network,the effectiveness of the proposed model is verified by two evaluation indicators:automatic evaluation and manual evaluation.The results show that the dialog model proposed in this paper has better performance of dialog generation,and can generate user-related replies,users are more willing to interact with them.The proposed dialog model improves the user experience.Finally,aiming at the problems such as huge storage burden of user model information and the difficulty of acquiring corpus in personalized dialog system,a personalized dialog system for service robots is designed based on the overall architecture of service robots cloud platform.Then the dialog service is deployed to the cloud.The validity of the proposed personalized dialog system is verified on the home service robot platform independently developed in the laboratory.