Knowledge Graph Embedding with 3D Compound Geometric Transformations

TL;DR

This paper introduces CompoundE3D, a family of knowledge graph embedding models using 3D geometric transformations like translation, rotation, scaling, reflection, and shear, demonstrating improved link prediction performance.

Contribution

It proposes a novel family of KGE models based on 3D compound geometric transformations, extending previous 2D and 3D models, with multiple variants and ensemble strategies.

Findings

CompoundE3D outperforms existing models on benchmark datasets.

Multiple variants of CompoundE3D capture diverse relation characteristics.

Ensemble of variants yields superior link prediction accuracy.

Abstract

The cascade of 2D geometric transformations were exploited to model relations between entities in a knowledge graph (KG), leading to an effective KG embedding (KGE) model, CompoundE. Furthermore, the rotation in the 3D space was proposed as a new KGE model, Rotate3D, by leveraging its non-commutative property. Inspired by CompoundE and Rotate3D, we leverage 3D compound geometric transformations, including translation, rotation, scaling, reflection, and shear and propose a family of KGE models, named CompoundE3D, in this work. CompoundE3D allows multiple design variants to match rich underlying characteristics of a KG. Since each variant has its own advantages on a subset of relations, an ensemble of multiple variants can yield superior performance. The effectiveness and flexibility of CompoundE3D are experimentally verified on four popular link prediction datasets.