Multi-Relational Semantic Awareness for Knowledge Graph Completion

With the advent of the data era and the wide application of knowledge graph technology, the integrity of knowledge graph has also attracted much attention. The real world is a heterogeneous network graph full of complex relationships. Existing knowledge representation learning to interact with entities and relationships improves the accuracy of knowledge characterization. However, it cannot mine the multi-layered potential semantic information under complex relationships. In order to solve this problem, this study proposes a multi-relationship semantic-aware knowledge graph completion method (MR-SAGCN), which enhances the expressive ability of knowledge triples and the generalization ability of the model by sensing the multi-layer semantic information between entities and the semantic information between relationships under different relationships, and adopting a multi-view decoder approach to enable the model to learn richer representations from different perspectives. Compared with the traditional knowledge representation learning model, our method shows corresponding improvements in various metrics on the generalized datasets FB15k-237 and WN18RR, which proves the effectiveness of our model.