Framework Design And Key Technical Research Of Customer Service Robot Based On Intention Recognition

Chatbot is an important research direction in the field of natural language processing,aiming to enable users to communicate with machines in the way of natural language.The question and answer system based on common question and answer pairs is an important realization method of chatbot.By comparing the similarity degree between the user's question and the question in the question and answer pairs,an accurate and concise answer is quickly returned,in which a comprehensive and accurate question and answer pair is the foundation of the question and answer system as well as the development bottleneck.In the field of customer service,a large number of real and artificial customer service data provides a very appropriate application scenario for the use of the question and answer system,and the accurate and efficient question and answer system also saves the cost of customer service for enterprises and improves the customer service efficiency.Based on the e-commerce customer service in the maternal and infant industry,this thesis has mainly been accomplished the following work:Firstly,aiming at the shortcomings of traditional FAQ framework,a new customer service robot framework is designed by adding intention recognition module on the basis of traditional FAQ framework.Then,in this paper,several key technologies in the framework of customer service robots,such as automatic keywords extraction,intention recognition and semantic similarity calculation,are studied in depth in combination with the practical application scenarios in this paper.A variety of keyword features are designed,and an automatic keyword extraction model is constructed by using XGBOOST algorithm.;an intention classification template and multiple intention recognition features are designed,and a multi-classification model is established by using Softmax algorithm;a variety of semantic similarity features are designed,and a semantic similarity calculation model is constructed based on XGBOOST algorithm.Finally,the three models established in this paper are tested and evaluated.Compared with traditional unsupervised method and supervised method,the accuracy rate of keyword automatic extraction method proposed in this paper has been greatly improved..The average accuracy rate of intention recognition is 78.4%.Compared with semantic similarity computing model based on TF-IDF,the accuracy rate of semantic similarity computing model based on automatic keyword extraction model is improved under different similarity thresholds.The three models are applied to the proposed customer service robot framework and the traditional FAQ framework.The experimental results show that the accuracy of the proposed framework is higher than that of the traditional FAQ framework.