Content recommendation method and system based on knowledge graph

Abstract

The invention discloses a content recommendation method based on a knowledge graph. The method comprises steps of obtaining the historical content clicked by a user in a set period of time, and determining a plurality of candidate contents similar to the historical content; for the historical content and the candidate content, respectively fusing semantic representation and knowledge representation of the historical content and the candidate content by using a KCNN to obtain a KCNN mapping result corresponding to each content; according to the KCNN mapping result of the historical content andthe candidate content, the score of each candidate content is determined through an Attention mechanism, and the score of each candidate content represents the probability that the candidate content is clicked by the user; and recommending the N candidate contents with the highest score to the user, where N is greater than or equal to 1. The invention further discloses a corresponding content recommendation system. The method is advantaged in that the potential interested knowledge content of a user can be fully mined, and more intelligent personalized content recommendation is realized.

Description

technical field [0001] The present application relates to the technical field of intelligent recommendation, in particular to a content recommendation method and system based on a knowledge graph. Background technique [0002] The main significance of the recommendation system is: in the era of information explosion, how to select appropriate information from a large amount of data to recommend to personalized users. Recommender systems hold great promise in the field of journalism, where, in general, journalism language is highly condensed, full of knowledge entities and commonsense knowledge. The current news recommendation method mainly relies on statistical machine learning, judges the similarity between news through news keywords, and then recommends similar news to users. [0003] An existing implementation is to recommend news based on content similarity. News recommendation based on content similarity, as the name suggests, is to recommend news content similar to t...

AI Technical Summary

Problems solved by technology

[0011] The above-mentioned news recommendation based on content similarity is limited to simple word matching, cannot be reasonably expanded, and has great inaccuracy

Moreover, the above method does not take into account the diversity of user interests. When the number of users reaches a certain order of magnitude, the entire recommendation system cannot provide personalized recommendations for users, thus losing the meaning of the recommendation system.

In addition, from a technical point of view, the inventors of this application found that: the existing news recommendation method does not use external knowledge, nor does it fully discover the potential knowledge-level connections between news, so it is impossible to tap the potential knowledge of interest to users

Images