Concept Commons Enhanced Knowledge Graph Representation

Yashen Wang, Xiaoye Ouyang, Xiaoling Zhu, Huanhuan Zhang

Abstract: Knowledge graphs (KGs) are regarded as important resources for a variety of artificial intelligence (AI) and auxiliary decision tasks but suffer from incompleteness. To address this challenge, a number of knowledge graph representation (KGR) and knowledge graph completion (KGC) methods have been developed using graph embeddings manners. Most existing methods focus on the structured information of native triples in encyclopaedia KG and maximize the likelihood of them. However, they neglect semantic commons contained in lexical KG. Recent researches aims at investigating the general framework for enhancing KGR or KGC methods with semantic signals (e.g., entity types, entity descriptions etc.,). However, their work almost modeled the semantic resources by an implicit way, leading their results lacked interpretability. To overcome this drawback, we propose a novel Concept Commons enhanced Knowledge Graph Representation model (named as C\(^2\)KGR), that integrates the structured information in encyclopaedia KG and the entity’s concepts in lexical KG, via both implicit manner and explicit manner. Experimental results demonstrate the efficiency and interpretability of the proposed model on the real-world datasets.