QuatE-D: A Distance-Based Quaternion Model for Knowledge Graph Embedding

TL;DR

QuatE-D introduces a distance-based scoring function for quaternion-valued KGEs, rotating head embeddings by a normalized relation quaternion and scoring via Euclidean distance to the tail. The approach leverages quaternion algebra (Hamilton product) to capture symmetry, antisymmetry, inversion, and composition, with unit-norm constraints and a margin-based loss. Empirical results on WN18, FB15k, WN18RR, and FB15k-237 show competitive MR and strong MRR/Hit@K, especially in MR reduction, while remaining more parameter-efficient than some baselines. The work highlights the interpretability and effectiveness of distance-based scoring in quaternion embeddings and points to future integrations with other quaternion-based models and dynamic graph settings.

Abstract

Knowledge graph embedding (KGE) methods aim to represent entities and relations in a continuous space while preserving their structural and semantic properties. Quaternion-based KGEs have demonstrated strong potential in capturing complex relational patterns. In this work, we propose QuatE-D, a novel quaternion-based model that employs a distance-based scoring function instead of traditional inner-product approaches. By leveraging Euclidean distance, QuatE-D enhances interpretability and provides a more flexible representation of relational structures. Experimental results demonstrate that QuatE-D achieves competitive performance while maintaining an efficient parameterization, particularly excelling in Mean Rank reduction. These findings highlight the effectiveness of distance-based scoring in quaternion embeddings, offering a promising direction for knowledge graph completion.

Figures (1)

• Figure 1: Impact of Embedding Dimension on Triple Classification Performance