A Recommendation Method Based on Heterogeneous Propagation Collaborative Knowledge Awareness Network

Abstract

The present invention proposes a recommendation method based on heterogeneous propagation collaborative knowledge perception network, which explicitly encodes collaborative signals between users and items through collaborative propagation, and proposes a method for combining collaborative signals with knowledge associations. Specifically, adopting a heterogeneous propagation strategy to explicitly encode the two kinds of information and applying a knowledge-aware attention mechanism to distinguish the contributions of different knowledge-based neighbors provides a new way to combine synergistic information with knowledge information .

Description

technical field [0001] The invention belongs to the technical field of personalized recommendation, and more particularly, relates to a recommendation method based on a heterogeneous propagation collaborative knowledge perception network. Background technique [0002] Recommender systems based on collaborative filtering usually have data sparsity and cold-start problems. The effective way to solve the above problems is to introduce auxiliary information. As a kind of knowledge auxiliary information, knowledge graph has been widely studied and applied in recommender systems. Existing knowledge graph-based recommendation methods mainly focus on how to effectively encode knowledge associations in knowledge graphs, but ignore the potential collaborative signals in user-item interaction information. [0003] The current recommendation method that combines collaborative signals with knowledge connections is to combine user-item interaction graphs and knowledge graphs to form coll...

AI Technical Summary

Problems solved by technology

[0006] 3. This method tends to bring more noise when constructing high-order user-item interactions, resulting in inefficient learning of latent vectors

Images