A self-supervised method for learning path-augmented knowledge graph embedding

doi:10.1016/j.engappai.2025.112315

A self-supervised method for learning path-augmented knowledge graph embedding

2025-09-20

0

PRE

AI

DOI：10.1016/j.engappai.2025.112315

Original

Original request for help

Save

Abstract

En 中文

Knowledge graphs (KGs) consist of factual triples that describe relations between entities in the real world. Knowledge graph embedding (KGE) aims to map entities and relations into constantly low-dimensional vectors, which is important for lots of downstream tasks (e.g., KG completion and information retrieval). Current KGE methods primarily rely on explicit structural patterns, neglecting latent contextual semantics behind those structures and resulting in sub-optimal performance. While some methods incorporate additional data (e.g., textual descriptions), such dependencies limit applicability due to additional data requirements. Furthermore, most KGE models suffer from limited supervision with sparse labeled triples, restricting their capacity to learn comprehensive semantic features.

A self-supervised method for learning path-augmented knowledge graph embedding

Abstract

Journal

Organization