Block-Diagonal Orthogonal Relation and Matrix Entity for Knowledge Graph Embedding
This work addresses a domain-specific problem for knowledge graph completion, offering an incremental improvement over existing rotation-based methods.
The paper tackled the challenges of limited flexibility and generalization in rotation-based knowledge graph embedding methods by introducing OrthogonalE, a model using matrices for entities and block-diagonal orthogonal matrices with Riemannian optimization for relations, which significantly outperformed state-of-the-art models and reduced relation parameters.
The primary aim of Knowledge Graph embeddings (KGE) is to learn low-dimensional representations of entities and relations for predicting missing facts. While rotation-based methods like RotatE and QuatE perform well in KGE, they face two challenges: limited model flexibility requiring proportional increases in relation size with entity dimension, and difficulties in generalizing the model for higher-dimensional rotations. To address these issues, we introduce OrthogonalE, a novel KGE model employing matrices for entities and block-diagonal orthogonal matrices with Riemannian optimization for relations. This approach enhances the generality and flexibility of KGE models. The experimental results indicate that our new KGE model, OrthogonalE, is both general and flexible, significantly outperforming state-of-the-art KGE models while substantially reducing the number of relation parameters.