HiHo: A Hierarchical and Homogenous Subgraph Learning Model for Knowledge Graph Relation Prediction

Relation prediction in knowledge graphs (KGs) aims to anticipate the connections between entities. While both transductive and inductive models are incorporated for context comprehension, we need to focus on two primary issues. First, these models only collate relations at each layer of the subgraph, overlooking the potential sequential relationship between different layers. Second, these methods overlook the homogeneity of subgraphs, thus impeding their ability to effectively learn the importance of relationships within the subgraphs. To address this challenge, we propose a hierarchical and homogenous subgraph learning model for KG relation prediction (HiHo). Specifically, we adopt a subgraph-to-sequence mechanism to learn the potential semantic associations between layers in the subgraph of a single entity, and thus model the hierarchy of the subgraph. Then, we implement a common preference inference mechanism that assigns higher weights to co-occurrence relations while learning the importance of each relation in the subgraphs of two entities, and thus models the homogeneity of the subgraph. In our study, we sequentially employ induction on each layer of subgraphs pertaining to the two entities for relation prediction. To assess the efficacy of our method, we perform experiments on five publicly available datasets. The results of our experiments demonstrate that our method surpasses the current state-of-the-art baselines in both transductive and inductive settings.