The Automatic Question Answering Customer Service System Of University Informationization

With the rapid development of the Internet,the process of universities informationization is getting faster and faster.There are a large number of users of various information systems in colleges and universities.However,due to the limitations of human resources and other aspects,customer service is still at the level of help documents and limited manual telephone consultation,which cannot meet the needs of teachers and students for timeliness and effectiveness of customer service.Aiming at the above problems,an automatic question-and-answer customer service system in the field of university informatization is designed and implemented to provide teachers and students efficient services such as question consultation,diagnosis and business handling,which mainly include modules such as question matching,knowledge base management,question classification and multiple rounds of dialogue.Among them,questions matching module by collecting network letter service domain corpus,combining with the Chinese wikipedia general corpus,using Word2 vec Skip-gram model,implemented algorithm for calculating the similarity of questions based on the word vector,to solve the problem that the semantic of question cannot be reflect in the calculation of similarity.Knowledge base management module through the analysis of the customer service demand of informatization in colleges and universities,the artificial classification of the existing questions,the design of the different categories of questions of knowledge description,formed a knowledge base that can be used in the automatic question and answer customer service system,realized the management of knowledge and training models,so as to solve the problems of unstructured knowledge in the field of information service in colleges and universities,high cost of artificial learning and late update.The problem of full-text matching of questions is solved by extracting the features of questions and classifying the questions with Support Vector Machine classifier.The rules-based multi-round dialog module combines relevant knowledge in the knowledge base to solve the problem that business processing and problem diagnosis need to interact with users.As a test set,word vectors trained with different corpora are used for similarity calculation respectively under two similarity algorithms.The experiment shows that the similarity calculation algorithm based on word co-occurrence has high accuracy.Compared with the Support Vector Machine classification results using different question features,the classification results based on tf-idf word bag features are better in the existing data set.By testing the system function,the feasibility of knowledge base and multi-round dialog strategy is verified.The result of the system performance test shows that the system can provide the question answering service for users in a timely,efficient and accurate manner.