Dynamic recommendation method and system based on knowledge graph embedding

Abstract

The invention discloses a dynamic recommendation method and system based on knowledge graph embedding. The method comprises the following steps: receiving an input query fact, searching whether the query fact exists in a knowledge graph, if the query fact exists, directly outputting a recommendation result, and if the query fact does not exist, updating the knowledge graph and then outputting therecommendation result; wherein the step of updating the knowledge graph is a dynamic knowledge graph embedding method combining a graph convolutional neural network and an ANALOGY model. The redundantlearning process of the knowledge graph can be reduced, so that the whole knowledge graph can be quickly updated when a user updates data or has a new tendency and favor each time, and the reliability and stability of whole dynamic recommendation are greatly improved.

Description

technical field [0001] The invention belongs to the technical field of information processing, and more specifically relates to a dynamic recommendation method and system based on knowledge map embedding. Background technique [0002] Knowledge graphs are very useful for recommendation systems. Current research on knowledge graphs for prediction and recommendation mainly focuses on the embedding methods of knowledge graphs. Knowledge graph embedding can convert complex heterogeneous directed graphs into low-dimensional vectors or linear transformations that meet certain characteristics. Traditional methods can generally be divided into translation distance method, matching semantic method and neural network method. The knowledge graph embedding algorithms used in traditional recommender systems have the following two shortcomings. [0003] (1) Since the entire knowledge graph must be relearned and embedded every time the knowledge graph is embedded, in actual use, the reco...

AI Technical Summary

Problems solved by technology

[0003] (1) Since the entire knowledge graph must be relearned and embedded every time the knowledge graph is embedded, in actual use, the recommendation system inevitably needs to update the user's data in real time and dynamically, while the traditional method obviously It is very cumbersome and inefficient for dynamic updates, and re-embedding the entire knowledge graph for each update will greatly affect the user experience and the effectiveness of the entire recommendation system

[0004] (2) For some data with clear rules and a small amount of data, the knowledge map embedding lacking a large amount of data learning lacks accuracy, often makes rash and wrong judgments, and cannot effectively use existing knowledge, thus affecting users' perception of the recommendation system. to experience

Images