Announcing DBpedia Release 2025-06

Announcements
, , , ,
1

Announcing DBpedia Release 2025-06

Copyright (C) 2025 OpenLink Software support@openlinksw.com

OpenLink Software is pleased to announce the release of a new DBpedia Knowledge Graph instance based on the DBpedia Databus System developed by the DBpedia Association.

This release represents a major evolution from the DBpedia 2022-12 dataset that has long served as the nucleus of the Linked Open Data (LOD) Cloud. With improved data consistency, richer entity descriptions, and tighter schema alignment across multiple domains, DBpedia 2025-06 strengthens its position as one of the most contextually rich and openly available structured data sources on the Web.

Service Endpoint Examples

We invite everyone to explore this new DBpedia Knowledge Graph and share feedback through our communication channels

DBpedia in the Age of AI

As the world shifts toward AI-driven information systems, the role of structured, contextualized data has become more critical than ever. Large Language Models (LLMs) and AI Agents increasingly rely on external knowledge graphs to provide grounding, disambiguation, and factual continuity—capabilities that pure text-based training data cannot guarantee.

DBpedia, as one of the longest-running and most interconnected open knowledge graphs, is ideally positioned to serve as a context provider for AI systems. It enables:

  • Entity Grounding and Linking — AI Agents can map ambiguous natural language terms (“Paris,” “Mercury,” “Jaguar”) to precise, dereferenceable DBpedia IRIs, ensuring that reasoning and retrieval operations reference real-world entities rather than vague text tokens.
  • Context Enrichment — DBpedia’s semantic links (e.g., dbo:influencedBy, dbo:spouse, dct:subject) supply contextual edges that enable LLMs and agents to infer meaning beyond text, supporting multi-hop reasoning and improved summarization or dialogue coherence.
  • Multimodal Integration — DBpedia’s typed relationships and URIs act as semantic anchors that can unify structured and unstructured sources—text, tabular data, and media—across APIs, vector databases, and enterprise systems.
  • Interoperability and Open Standards — By adhering to W3C and LOD standards, DBpedia can be seamlessly incorporated into AI Agent architectures compliant with MCP (Model Context Protocol), A2A (Agent-to-Agent), and OpenAPI. This allows agents to query, reason, and exchange knowledge using uniform semantics.

VOID-based Metadata Graph

Description DBpedia 2022-12 DBpedia 2025-06
Number of triples 1,152,980,375 1,320,461,984
Distinct subjects 55,792,426 . 49,758,557
Distinct properties 54,812 56,392
Distinct objects 273,601,360 194,373,746
Distinct classes 483,639 -
Entities 42,075,815 33,286,806
sameAs links 136,614,458 136,721,988
seeAlso links 302,408 356,036

Related