Research On Customer Service Scheduling And Conversation Topic Extraction Technology In WeChat-based Customer Service System

With the development of Mobile Internet and the popularization of smartphones,WeChat has great value in communication,promotion,marketing and promotion,it occupies an increasingly important position in people's lives by its own advantages and possessing a large user base.The customer service system plays an important role in understanding customer needs,solving customer problems,grasping market orientation and enhancing corporate image,as the main way for enterprises to communicate with customers.Because traditional online automatic question-answering system and instant messaging-based customer service system have some problems,such as wasting human resources,difficult sharing of resources and low accuracy.Developing a new generation of customer service system with universality,versatility,personalization and intelligence for customers and businesses is of great significance.Therefore,according to the characteristics of customer service system and WeChat,this paper designs and develops a new intelligent WeChat-based customer service system.Under the project,the research on customer service scheduling and text subject extraction is carried out as the following:Aiming at the reducing of chat text words irregular random words and sparse text features in the process of instant messaging,this paper proposes a customer service scheduling method based on gated recurrent unit(GRU)real-time emotion analysis.This method firstly combines the characteristics of WeChat chat text,uses GRU model to analyze real-time emotions of users,and then classifies users according to the changes of users' emotions and problems.Finally,dispatches the corresponding customer service categories for service.The effectiveness of this method is verified through experiments,which not only can grasp the emotional changes of users in real time,but also enable users to experience high quality and efficient professional services.In view of the problems of the traditional theme model such as the high dimension,the weakness of the intelligibility,the ambiguous boundaries of the various themes and the interweaving of the theme,the paper proposes a topic-oriented extraction method for the short text of conversation.Firstly,LDA theme model is usedto get the topic distribution in the target text in different time periods.Then the theme vector is constructed based on the topic distribution.Then,the sparse self-encoder is used to reduce the dimension of the theme vector and extract the feature.At last,the similarity of the extracted theme feature vector Compare sorts to get a more expressive theme.Mining user preference features and extracting user chat topics can provide precise knowledge of marketing,training customer service and other aspects of the reserve,which is an important part of diversifying the smart customer service system and other systems.Experiments show that this method can extract topics that can cover the main information of the text and have strong explanatory power.Finally,we designed and implemented the WeChat intelligent customer service system,and applied the customer service scheduling and thematic extraction technology to the system according to the above method model,so that the WeChat customer service system designed in this paper has made further progress toward diversification,personalization and intelligence step.