Similarity Calculation Algorithm for Intelligent Electronic Customer Service Problems

Abstract

Customer service question similarity calculation is a key technology in the development of intelligent electronic customer service system, and its main challenge comes from how to find the correlation between words. In response to this challenge, the paper proposes a weighted text similarity calculation method based self-attention. The paper uses deep neural network BiLSTM and self-attention vectors to carry out text similarity research, and explores the problem of deep neural network and self-attention vectors describing the semantic relationship between words. The research of this paper is of great significance to improve the service quality of the intelligent electronic customer service system, and provides new ideas for the accurate comparison of text semantics. The paper takes the semantically vague and complex Chinese text as the research object, and uses the deep neural network structure design, self-attention vector calculation, dynamic weighting, and synchronization comparison as processing methods. Combined with the development of the Weizhong bank intelligent electronic customer service system, an intelligent electronic Customer service question matching experiment platform is constructed. Experimental results show that the dynamic weighted self-attention text similarity calculation model is superior to the existing models in terms of calculation accuracy and running time. It has certain reference value for the development of similar intelligent electronic customer service systems.