Interpretable recommendation system based on knowledge graph

Abstract

The invention discloses an interpretable recommendation system based on a knowledge graph. The system comprises a sequence knowledge graph construction module which extracts predefined entities and relationships from a purchase sequence of a user and user comments; a user representation learning module which is used for designing a graph attention network to learn the short-term preference of the user, redesigning the structure of a gating loop unit so as to fuse the long-term preference and the short-term preference of the user, and obtaining a user vector representation which captures the dynamic nature of the user preference; an article representation learning module which is used for designing a self-attention layer to learn user specific vector representation of the article; and a score prediction module which is used for combining the vector representation of the user and the vector representation of the article to obtain a predicted score of the article by the user. According to the system, the method achieves the purposes of high prediction precision and high-efficiency explanation, and has high practical value and popularization value.

Description

technical field [0001] The invention belongs to the technical field of knowledge graphs, and in particular relates to an interpretable recommendation system based on knowledge graphs. Background technique [0002] In recent years, explainable recommender systems have received more and more attention in academia and industry. Related studies have shown that explainable recommender systems can not only improve users' acceptance of recommended items, but also improve the transparency and persuasion of recommender systems. power, efficiency, reliability and user satisfaction. [0003] In order to improve the interpretability of the recommendation system, most of the existing models use user comments containing rich semantic information as the input of the model. Although the interpretability of the model has been improved, these explainable recommendation models still have some limitations. First of all, most of these explainable recommendation models do not take the dynamics o...

AI Technical Summary

Problems solved by technology

Secondly, the comment-level and feature-level explanations provided by these interpretable recommendation models have limitations. The comment-level explanation is to provide the entire comment of other users to the current user as an explanation. Since user comments contain a lot of personal preference information , so that the explanation at the comment level is too redundant, and the explanation at the feature level is to provide an item feature that the current user is concerned about as an explanation. Because a feature word contains too little information, the explanation at the feature level is too concise, which may can confuse users

Finally, although these recommendation models all use user reviews as input to the model, none of these recommendation models deeply mine the rich semantic information contained in reviews.

Images