Improving Inductive Link Prediction Using Hyper-Relational Facts

TL;DR

This paper explores how hyper-relational knowledge graphs, which include richer information like qualifiers, can improve inductive link prediction tasks, especially on unseen entities, using graph neural networks.

Contribution

It introduces a classification of inductive settings for hyper-relational KGs and demonstrates performance gains over triple-only baselines in semi- and fully inductive scenarios.

Findings

Qualifiers improve link prediction accuracy by up to 6% on Hits@10.

Hyper-relational KGs outperform triple-based models in inductive tasks.

New benchmark datasets for hyper-relational inductive link prediction.

Abstract

For many years, link prediction on knowledge graphs (KGs) has been a purely transductive task, not allowing for reasoning on unseen entities. Recently, increasing efforts are put into exploring semi- and fully inductive scenarios, enabling inference over unseen and emerging entities. Still, all these approaches only consider triple-based \glspl{kg}, whereas their richer counterparts, hyper-relational KGs (e.g., Wikidata), have not yet been properly studied. In this work, we classify different inductive settings and study the benefits of employing hyper-relational KGs on a wide range of semi- and fully inductive link prediction tasks powered by recent advancements in graph neural networks. Our experiments on a novel set of benchmarks show that qualifiers over typed edges can lead to performance improvements of 6% of absolute gains (for the Hits@10 metric) compared to triple-only baselines. Our code is available at \url{https://github.com/mali-git/hyper_relational_ilp}.