Towards Semantically Enriched Embeddings for Knowledge Graph Completion
Mehwish Alam, Frank van Harmelen, Maribel Acosta
TL;DR
The paper surveys semantic enrichment for knowledge graph completion, arguing that traditional KG embeddings largely ignore TBox and external knowledge. It tracks evolution from purely structural, transductive approaches to methods that incorporate type hierarchies, description-logic axioms, and LLMs, while critically examining their limitations. It highlights evaluation challenges, including dataset biases and the gap between closed-world metrics and real-world applicability, and proposes directions toward semantic embeddings, external resources, and more representative benchmarks. Overall, it lays out a roadmap for integrating ontological semantics and external knowledge into KG completion with practical recommendations for datasets, evaluation, and modeling approaches.
Abstract
Embedding based Knowledge Graph (KG) Completion has gained much attention over the past few years. Most of the current algorithms consider a KG as a multidirectional labeled graph and lack the ability to capture the semantics underlying the schematic information. In a separate development, a vast amount of information has been captured within the Large Language Models (LLMs) which has revolutionized the field of Artificial Intelligence. KGs could benefit from these LLMs and vice versa. This vision paper discusses the existing algorithms for KG completion based on the variations for generating KG embeddings. It starts with discussing various KG completion algorithms such as transductive and inductive link prediction and entity type prediction algorithms. It then moves on to the algorithms utilizing type information within the KGs, LLMs, and finally to algorithms capturing the semantics represented in different description logic axioms. We conclude the paper with a critical reflection on the current state of work in the community and give recommendations for future directions.