NE-GCN: Advancing Knowledge Graph Link Prediction with Node2vec-Enhanced Graph Convolutional Networks

Knowledge graphs (KGs) play a vital role in enhancing search results and recommendation systems. With the rapid increase in the size of the KGs, they are becoming inaccuracy and incomplete. This problem can be solved by the knowledge graph completion methods. In this paper we use a novel method for knowledge graph link prediction named Node2vec Enhanced Graph Convolutional Network (NE-GCN), for computing pairwise occurrences of entity-relation pairs in the dataset to construct a joint learning model. Given a knowledge graph, NE-GCN constructs a single graph considering entities and relations as individual nodes. NE-GCN then computes weights for edges among nodes based on the pairwise occurrence of entities and relations. Next, uses Graph Convolution neural Network (GCN) to update vector representations for entity and relation nodes. This work opens up