Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Survey on Knowledge Graph Structure and Knowledge Graph Embeddings

Jeffrey Sardina, John D. Kelleher, Declan O'Sullivan

TL;DR

This survey addresses how Knowledge Graph structure and KGEM hyperparameters jointly influence link prediction ($LP$) performance, highlighting biases driven by degree distribution and topological imbalance. It consolidates findings from numerous studies across structure metrics, hyperparameter choices, and benchmark KGEM evaluations to reveal consistent effects of node degree, centrality, and relation frequencies on learning and evaluation. A key takeaway is that hyperparameter preference is context-dependent, varying with both the KG and KGEM, and that many reported gains may be due to structure-induced biases rather than intrinsic model improvements. The authors advocate for structure-aware evaluation, ontological considerations, and structurally controlled benchmarks to advance holistic understanding and fair comparisons in KGEM research.

Abstract

Knowledge Graphs (KGs) and their machine learning counterpart, Knowledge Graph Embedding Models (KGEMs), have seen ever-increasing use in a wide variety of academic and applied settings. In particular, KGEMs are typically applied to KGs to solve the link prediction task; i.e. to predict new facts in the domain of a KG based on existing, observed facts. While this approach has been shown substantial power in many end-use cases, it remains incompletely characterised in terms of how KGEMs react differently to KG structure. This is of particular concern in light of recent studies showing that KG structure can be a significant source of bias as well as partially determinant of overall KGEM performance. This paper seeks to address this gap in the state-of-the-art. This paper provides, to the authors' knowledge, the first comprehensive survey exploring established relationships of Knowledge Graph Embedding Models and Graph structure in the literature. It is the hope of the authors that this work will inspire further studies in this area, and contribute to a more holistic understanding of KGs, KGEMs, and the link prediction task.

A Survey on Knowledge Graph Structure and Knowledge Graph Embeddings

TL;DR

This survey addresses how Knowledge Graph structure and KGEM hyperparameters jointly influence link prediction () performance, highlighting biases driven by degree distribution and topological imbalance. It consolidates findings from numerous studies across structure metrics, hyperparameter choices, and benchmark KGEM evaluations to reveal consistent effects of node degree, centrality, and relation frequencies on learning and evaluation. A key takeaway is that hyperparameter preference is context-dependent, varying with both the KG and KGEM, and that many reported gains may be due to structure-induced biases rather than intrinsic model improvements. The authors advocate for structure-aware evaluation, ontological considerations, and structurally controlled benchmarks to advance holistic understanding and fair comparisons in KGEM research.

Abstract

Knowledge Graphs (KGs) and their machine learning counterpart, Knowledge Graph Embedding Models (KGEMs), have seen ever-increasing use in a wide variety of academic and applied settings. In particular, KGEMs are typically applied to KGs to solve the link prediction task; i.e. to predict new facts in the domain of a KG based on existing, observed facts. While this approach has been shown substantial power in many end-use cases, it remains incompletely characterised in terms of how KGEMs react differently to KG structure. This is of particular concern in light of recent studies showing that KG structure can be a significant source of bias as well as partially determinant of overall KGEM performance. This paper seeks to address this gap in the state-of-the-art. This paper provides, to the authors' knowledge, the first comprehensive survey exploring established relationships of Knowledge Graph Embedding Models and Graph structure in the literature. It is the hope of the authors that this work will inspire further studies in this area, and contribute to a more holistic understanding of KGs, KGEMs, and the link prediction task.

Paper Structure

This paper contains 19 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Measures of Knowledge Graph Structure
  3. Rossi et al. (2020)
  4. Sadeghi et al. (2021)
  5. Bonner et al. (2022)
  6. Zietz et al. (2024)
  7. Mohamed et al. (2020)
  8. Kotnis et al. (2017)
  9. Rossi et al. (2021)
  10. Sardina et al. (2022)
  11. Sardina et al. (2024)
  12. Dave et al. (2024)
  13. Hyperparameter Preference and Link Prediction Performance
  14. Kotnis et al. (2017)
  15. Sameh et al. (2019)
  16. ...and 4 more sections

Figures (1)

  • Figure 1: An overview of hyperparameter and graph structure influence on KGEM performance on the link prediction task, annotated by evidence in the literature. A directed edge from a subject node to an object node indicates that the item subject is dependent upon the object. Nodes are connected to Link Prediction Performance if they have been shown to ave concrete influence on overall link prediction performance. Note that as the literature has shown that hyperparameters are generally dependent on both the KGEM and KG used dark-into-lightold-dog-new-tricks, only hyperparameters that have been specifically studied are included.