Research And Application Of Key Algorithm Of Intelligent Customer Service Answering System

With the rapid development of information technology,the demand of online customer service is more and more big,the artificial customer service is difficult to meet the needs of the present mass QA And intelligent customer service system based on machine learning,to the understanding question the accuracy of the retrieval results of quickness,good adaptive for big data QA scene,has the very high practical and research value This paper deeply studies the key technologies,intelligent customer service system for question classification technology and questions of similarity technology were studied,the main work has the following points:First,on the basis of K-means clustering and simple Bias classification algorithm,this paper proposes a NBKC algorithm.The algorithm optimizes the problem of low accuracy of K-means clustering algorithm and low efficiency of simple Bias classification algorithm.First,the samples are clustered by K-means,and the correct problem sets in the clustering results are used to train naive Bayes classifier,and the classifier is used to classify the wrong problem sets in the clustering results for two times.The experiment shows that the improved NBKC algorithm can not only quickly classify the query samples without classification,but also have higher classification accuracy.Secondly,on the basis of analyzing the characteristics of SimHash similarity,semantic similarity and word order similarity,a hybrid similarity algorithm combining three elements is proposed.First,we use the SimHash method to filter quickly,and retain the question answer pairs which are similar to the user input question in the FAQ library.Then we match the most similar questions through the semantic similarity and word order similarity,and return the corresponding answers to the users.Experiments show that the hybrid algorithm can take account of the accuracy and query time of the system to meet the user's query accuracy and speed.Finally,based on the above two improved algorithms,we implemented an intelligent customer service QA system based on the restricted domain from the following 4 aspects: requirement analysis,overall design,detailed design and system implementation.