Multi-dimensional approach to knowledge graph completion: integrating entity types, relations, and paths

Knowledge graphs (KGs), as highly structured information storage systems, play a crucial role in various fields. However, existing knowledge graphs still face the issue of incomplete knowledge, which limits their potential in practical applications. To address this problem, this paper proposes <strong><u>A</u></strong> novel <strong><u>M</u></strong>ulti-<strong><u>D</u></strong>imensional <strong><u>K</u></strong>nowledge <strong><u>G</u></strong>raph <strong><u>C</u></strong>ompletion method (AMD-KGC) that integrates entity types and topology-aware information. This method significantly improves the accuracy and generalization ability of knowledge graph completion by integrating entity type information, relation context information, and relation path information. Specifically, AMD-KGC learns from contextual information, uses an attention mechanism to adaptively calculate entity type weights, and introduces entity path information to further enrich the representation of entity pairs. Experimental results demonstrate that on the FB15k-237, WN18RR, and EDUDB (Educational Database Engine) datasets, AMD-KGC improves key performance metrics MRR (Mean Reciprocal Rank) and Hits@10 by 9.2% and 10.5%, respectively, compared to current leading technologies, demonstrating its superiority in knowledge graph completion tasks.