An intelligent customer service question matching method based on knowledge base self-learning

Abstract

The invention discloses an intelligent customer service question matching method based on knowledge base self-learning. The question matching method includes three parts: knowledge base construction, semantic representation learning and semantic vector matching; the knowledge base is composed of multiple similarities of different questions. Question sentence composition, semantic representation learning part, consists of input layer, semantic representation layer, similarity calculation layer, interval classification layer, this part uses classification task and margin-loss function to complete semantic vector representation learning and matching sorting, semantic vector matching Part, according to the semantic representation layer of the semantic representation learning part, extract the semantic vector of the user question and the semantic vector of the knowledge base, calculate the similarity, sort according to the similarity, and return the N knowledge base questions with the highest similarity as the final matching result . The invention greatly improves the accuracy rate of intelligent customer service question matching, realizes self-learning based on the knowledge base, reduces the workload of manual labeling, and improves the work efficiency and the reply satisfaction of the intelligent customer service.

Description

technical field [0001] The invention relates to the field of sentence matching, in particular to an intelligent customer service question matching method based on knowledge base self-learning. Background technique [0002] The core problem of query matching of retrieval-type intelligent customer service is how to carry out the vector representation of the question and how to retrieve the question most similar to the user's question from the massive knowledge base. At present, there are two widely used methods. One is to pre-train the word vector, segment the input question sentence, and then extract the keyword, and manually weight the word vector of the keyword according to the part of speech and word order of the keyword. The weighted word vector represents the input question, and the similarity is calculated with the question in the knowledge base, and the question with the highest similarity is returned as the final matching result. [0003] However, the problem of this...

AI Technical Summary

Problems solved by technology

The method of the present invention is based on character learning, without word segmentation, and at the same time eliminates the manual labeling link, uses the method of interval classification to solve the problem of matching and sorting, and improves accuracy while facilitating maintenance

Images