FB15k-CVT: A Challenging Dataset for Knowledge Graph Embedding Models - Mines Saint-Étienne Access content directly
Conference Papers Year : 2023

FB15k-CVT: A Challenging Dataset for Knowledge Graph Embedding Models

Abstract

Knowledge graphs (KGs) have become an essential component of neuro-symbolic AI research. A KG is a uniform source of information in which physical-world entities are represented as vertices of a directed edge-labeled graph. In the context of representation learning, edge labels of a KG are called relations, and its edges are called facts or triples [5]. KGs can be leveraged in a great variety of AI applications. Over the past decade, many KG Embedding Models (KGEMs) have been developed for that purpose [5]. By representing entities and relations as numeric structures in a vector space, KGEMs provide a way to integrate both symbolic and sub-symbolic knowledge, enabling efficient processing and reasoning over complex and heterogeneous data. Most KGEMs are evaluated against datasets that are derived from Freebase, a (now archived) public KG containing millions of entities and billions of facts.

Domains

Web
No file

Dates and versions

emse-04138834 , version 1 (23-06-2023)

Identifiers

  • HAL Id : emse-04138834 , version 1

Cite

Mouloud Iferroudjene, Victor Charpenay, Antoine Zimmermann. FB15k-CVT: A Challenging Dataset for Knowledge Graph Embedding Models. SeReCo Summer Workshop 2023, Jul 2023, Waischenfeld, Germany. ⟨emse-04138834⟩
40 View
0 Download

Share

Gmail Mastodon Facebook X LinkedIn More