Learning Hierarchy-Aware Knowledge Graph Embeddings for Link Prediction

Knowledge graph embedding, which aims to represent entities and relations as low dimensional vectors (or matrices, tensors, etc.), has been shown to be a powerful technique for predicting missing links in knowledge graphs. Existing knowledge graph embedding models mainly focus on modeling relation patterns such as symmetry/antisymmetry, inversion, and composition. However, many existing approaches fail to model semantic hierarchies, which are common in real-world applications. To address this challenge, we propose a novel knowledge graph embedding model--namely, Hierarchy-A ware Knowledge Graph E mbedding (HAKE)-- which maps entities into the polar coordinate system. HAKE is inspired by the fact that concentric circles in the polar coordinate system can naturally reflect the hierarchy. Specifically, the radial coordinate aims to model entities at different levels of the hierarchy, and entities with smaller radii are expected to be at higher levels; the angular coordinate aims to distinguish entities at the same level of the hierarchy, and these entities are expected to have roughly the same radii but different angles. Experiments demonstrate that HAKE can effectively model the semantic hierarchies in knowledge graphs, and significantly outperforms existing state-of-the-art methods on benchmark datasets for the link prediction task. 1 Introduction Knowledge graphs are usually collections of factual triples--(head entity, relation, tail entity), which represent human knowledge in a structured way. In the past few years, we have witnessed the great achievement of knowledge graphs in many areas, such as natural language processing (Zhang et al. 2019), question answering (Huang et al. 2019), and recommendation systems (Wang et al. 2018). Although commonly used knowledge graphs contain billions of triples, they still suffer from the incompleteness problem that a lot of valid triples are missing, as it is impractical to find all valid triples manually. Therefore, knowledge graph completion, also known as link prediction in knowledge graphs, has attracted much attention recently. Link prediction aims to automatically predict missing links between entities based on known links. It is a challenging task as we Equal contribution. Inspired by word embeddings (Mikolov et al. 2013) that can well capture semantic meaning of words, researchers turn to distributed representations of knowledge graphs (aka, knowledge graph embeddings) to deal with the link prediction problem.