An Intelligent Customer Service System for Securities Industry based on Compound Corpus and Text Vector

Abstract

Customer service is one of essential tasks for securities companies. The traditional manual service mode has disadvantages such as high labour intensity and poor customer service experience. This paper proposes an intelligent customer service system based on Fasttext and word vectors. By constructing the securities knowledge database, and using Fasttext to train the word vector model from the Chinese compound corpus, the performance of “sentence vector + cosine distance” and “word vector similarity matrix + CNN feature extraction” methods in calculating semantic similarity are analyzed based on single corpus and compound corpus, respectively. The results show that “sentence vector + cosine distance” has higher performance on accuracy rate.