Design And Implementation Customer Service System Based On Man-machine Dialogue Technology For Shopping Domain

In recent years,e-commerce has gradually become a mainstream shopping method.In particular,e-commerce platforms have become an extremely important sales channel for electronic products.It’ s necessary to improve the in-sale and after-sale customer service in e-commerce platforms.Via the task-based dialogue system,users can get response more conveniently;E-commerce platforms can lower human costs while making communication more convenient.In the field of electronic product shopping,existing customer service systems have some shortcomings.For example,the intelligent customer service cannot understand the user’s semantics precisely and therefore gives irrelevant responses,or the dialogue system is inefficient and requires a long inquire process to suit users’needs.In order to solve the above problems and bring users a better shopping service experience,this paper has done the following work:(1)Investigate the research background,significance and current status of the task-based dialogue system,and elaborate on the problems and pain points of the electronic product dialogue system in the current shopping e-commerce platform.Analyze the overall business requirements for the existing problems,and split them into modular processing,mainly split into functional modules and non-functional modules.The nonfunctional modules are mainly the front-end interaction module between the user and the system,and the functional module Contains natural language understanding module,dialogue management module and dialogue generation module on a given data set.(2)According to actual needs,considering that the data of this subject is mainly customer service data in the field of electronic products during the in-sale and after-sale,the overall design and detailed design of the system is carried out for the after-sale phase of the electronic product field.The overall framework content of the overall design system and the connections between the various sub-modules,the detailed design has elaborated and explained each sub-module in detail.In addition,this thesis also uses flowcharts,UML class diagrams and other methods to assist in introducing the detailed information of the module.(3)For the core algorithm module and natural language understanding module of the system,the intention-slot joint modeling is adopted,and the ERNIE-BiLSTM-CRF model is constructed on the basis of ERNIE-base.The experimental results show that the output layer of the ERNIE model is spliced The BiLSTM-CRF model achieves the best results,with an accuracy rate of 91.34%.On the given dataset,compared with the mainstream methods used in the current natural language understanding module,the accuracy of the module is improved,and the response of the final customer service system is more accurate.(4)Demonstrate and test the system,perform functional tests on several important sub-modules in the system,verify the accuracy of the model methods used in this topic through the evaluation indicators of each part,and verify the effectiveness of the functional modules of the dialogue system.Finally,a specific test case is used to show the results of the interaction between the user and the customer service system,which proves the usability and completeness of the shopping customer service system for the in-sale and after-sale stages of the electronic product field.