Integrating adaptive fuzzy embedding with topology and property hypergraphs: Enhancing membership degree-aware knowledge graph reasoning

Knowledge Graphs (KG) leverage the Resource Description Framework (RDF) to model clear and static resources, enabling the revelation of latent meanings through embedding and reasoning. However, their effectiveness diminishes when facing the inherent vagueness and dynamism of real-world information. To overcome this limitation, we introduce a novel adaptive fuzzy RDF KG embedding approach. This method represents fuzzy knowledge by integrating topological structures and descriptive properties while generating membership degrees to enhance the reasoning capabilities of KG. Specifically, our approach introduces a way to create embeddings based on local and global topological pattern mining. Then, path quality-based adaptive learning is implemented. Next, we incorporate entity descriptive information and construct a property hypergraph from another perspective to learn embeddings. By creatively combining these three components, our method can generate a fuzzy membership degree to aid knowledge reasoning. To thoroughly assess the method's effectiveness, experiments were conducted around four key questions, evaluating algorithm performance compared with other work and assessing knowledge reasoning effectiveness based on logical inference rules on Unmanned Aerial Vehicle (UAV) datasets.