Wikidata Truthy: Core Facts as RDF Triples for Machine Learning

Wikidata Truthy contains RDF triples from Wikidata, converted to HuggingFace dataset format for machine learning pipelines. The dataset includes preferred statements from Wikidata, comprising approximately 100 million entities and 2 billion triples. It was created by CleverThis and last updated in December 2025.

Use Cases

• Train knowledge graph embedding models based on the RDF triple structure.
• Perform entity linking and disambiguation using the core Wikidata facts.
• Build or augment knowledge bases for question-answering systems.
• Analyze relationships between entities within the Wikidata knowledge graph.

Strengths

• Contains approximately 2 billion triples, providing a large-scale knowledge base.
• Includes core facts from Wikidata, focusing on preferred statements for consistency.
• Converted to HuggingFace dataset format for easier integration into ML workflows.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2025-12-07 11:53:42; freshness should be verified.

Provenance

Source

https://dumps.wikimedia.org/wikidatawiki/entities/latest-truthy.nt.bz2

Collection Method

RDF triples from Wikidata Truthy converted to HuggingFace dataset format.

Freshness

Last updated 2025-12-07 11:53:42.