Cultural Resource Recommender System with Knowledge Graph Convolutional Network

Due to the abundance of cultural resources on the Internet, traditional information retrieval has not met people's requirements. Therefore, intelligent recommendation of cultural resources has become an valuable research. However, previous studies on recommender system fail to exploit the higher-order structural information in knowledge graphs fully. Therefore, this paper proposes an Improved Knowledge Graph Convolutional Network (KGCN-I) model, which expresses the semantic information in the knowledge graph by weighting the user's score on the relationship. We use a brand-new aggregation function to replace the original KGCN layer's aggregation function and expand each entity's receptive field in the knowledge graph by stacking multiple KGCN layers to capture the high-order personalized interests of users. We compare with the state-of-the-art baseline algorithms on two real datasets. Experiments show that our proposed model is superior to the baseline algorithms on the precision, recall and F1.