MuCo-KGC: Multi-Context-Aware Knowledge Graph Completion

MuCo-KGC: Multi-Context-Aware Knowledge Graph Completion

Authors: Gul H., Naim A. G., & Bhat A. A.

Journal: Pacific-Asia Conference on Knowledge Discovery and Data Mining

Tags: graph, bert, neural-model, link-prediction

Link: URL

Abstract:

Traditional knowledge graph completion methods often ignore the rich contextual information available in multi-relational graphs. We propose MuCo-KGC, a multi-context-aware approach that leverages diverse contextual signals to improve link prediction performance. Our method achieves state-of-the-art results on several benchmarks while maintaining computational efficiency.

Methodology:

MuCo-KGC incorporates multiple types of contextual information including entity neighborhoods, relation patterns, and temporal dynamics. We use attention mechanisms to weight different contexts and employ advanced negative sampling techniques. Evaluation is performed on FB15k-237, WN18RR, and custom biomedical KGs with standard metrics (MRR, Hit@k).