Research on Intelligent Recommendation and Information Organization Methods for Academic Resources Integrating Knowledge Graph

To address the increasingly severe problem of academic information overload and overcome the limitations of traditional recommendation systems in processing sparse and heterogeneous academic data, this paper proposes a knowledge graph based academic resource intelligent recommendation and information organization framework (KGFAR). This framework aims to build a comprehensive and unified academic knowledge graph (AcadKG) by integrating heterogeneous academic knowledge from multiple sources, and based on this, provide accurate personalized recommendations. Methodologically, this study first designed a hybrid entity alignment model that combines structure and attribute similarity for offline construction of highquality AcadKG. Subsequently, in the online recommendation stage, user interaction data is integrated into AcadKG to form a collaborative knowledge graph (CKG), and a graph neural network (GNN) model based on knowledge aware attention mechanism is applied on top of it. This model can adaptively learn the importance of different neighboring entities and relationships through multi-layer information propagation, thereby accurately capturing the high-order interests of users. The simulation experiment results show that compared to various baseline models, KGFAR has achieved significant improvement in recommendation accuracy. This study not only provides an effective solution for academic resource recommendation, but also offers new ideas for knowledge organization and discovery in complex information networks.