Wikipedia Knowledge Graph: 16.9 Million Ontologies Extracted via LLM

Jotschi's Wikipedia Knowledge Graph contains 16,958,654 extracted ontologies from a subset of English Wikipedia articles. The dataset was created by processing the English Wikipedia 20231101.en dataset through an LLM pipeline for fact and ontology extraction. It was last updated on December 3, 2024.

Use Cases

• Train or evaluate knowledge graph embedding models based on the extracted ontology relationships.
• Benchmark ontology extraction and linking algorithms against the large-scale, structured knowledge base.
• Enhance question-answering systems by providing a structured factual backbone derived from Wikipedia.
• Populate semantic search indices with machine-readable relationships between entities and concepts.

Strengths

• Contains 16,958,654 extracted ontologies, indicating a large-scale knowledge base.
• Derived from the English Wikipedia 20231101.en dataset, a well-known and widely used corpus.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but the specific content, structure, and potential biases of the extracted ontologies are not detailed.
• Freshness should be verified as the source Wikipedia snapshot is from November 2023.

Provenance

Source

Subset of the English Wikipedia 20231101.en dataset.

Collection Method

Created via an LLM pipeline: Wikipedia article → Chunking → Fact extraction → Ontology extraction.

Freshness

Last updated 2024-12-03 06:42:37