Abstract
Recently, graph neural networks (GNNs) have gained prominence in recommender systems (RS) due to their capability to extract vital features and understand intricate relationships. However, GNNs exhibit limitations in their ability to capture fine-grained semantics in a knowledge graph (KG) and are often insufficient in effectively modeling user-item interactions. One approach to address these limitations is personalized knowledge-aware recommendation. In this paper, a novel RS, called LGKAT, is proposed that uses a combination of user-item graph and knowledge graph. It allows for more precise and nuanced modeling of user-item interactions, aiding the recommender system in learning meaningful node representations. One of the contributions of the proposed method is to use a novel integration of light graph convolutional network (LightGCN) in RSs to efficiently manage common signals for user and item embeddings. Another novelty of this paper is to propose an efficient attention sub-network that encodes rich semantic meanings from the knowledge graph into refined item embeddings in a personalized manner. Extensive tests were conducted on four well-known datasets. The metrics of F1_score and recall is used for the evaluation of the proposed method. The experimental results show the significant superiority of the proposed method compared to the state-of-the-art methods. The obtained results show that the integration of LightGCN in personalized knowledge-aware recommendation systems can effectively tackle limitations of current recommender systems and improve the quality of recommendations.