A multitask learning recommendation approach fusing knowledge graph and user intention

Knowledge graph as a powerful auxiliary information to promote the development of recommender systems, but in the traditional knowledge graph feature learning its main purpose is to characterize the ternary, rather than to improve the recommendation performance, which results in the inability to significantly improve the effectiveness of the recommendation results. In order to improve the combination of knowledge graph and recommendation, effective learning in which the user's characterization of the item's intention. In this paper, we propose a recommendation method that combines knowledge graph feature learning and user intent to make recommendations with understanding of the actual intent of the user. Specifically, the users' intention for the project is understood in the interaction information between the user and the project, the user behavioral preference information is portrayed by the attention score, and the Pearson correlation coefficient is used to model the independence of the intention. Users and items are represented by path-aware convolutional neural networks, with a cross-learning module that combines the tasks of intent-based learning and knowledge graph feature learning, complementing user and item information with shared features to improve recommendation performance. Experimental results on two open-source datasets demonstrate the effectiveness of the method proposed in this paper.