NELL ontology and knowledge base

What it is This archived Linked Data server publishes the ontology and knowledge base produced by NELL, the Never-Ending Language Learning project in Carnegie Mellon’s Read-The-Web initiative. NELL has iteratively visited web pages and extracted knowledge from their unstructured content since January 2010, while NELL2RDF publishes that material as RDF.

Key concepts Release identifiers combine the NELL2RDF software version with the source NELL iteration, so 0.3#1100 means software 0.3 over iteration 1100. The server offered vanilla RDF plus provenance modeled through RDF reification, n-ary relations, named graphs, singleton properties, or NDFluents, distributed as compressed N-Triples or HDT and licensed CC0 1.0.

How you’d use it Download the ontology, NELL2RDF metadata ontology, and a dump matching the required provenance representation, or query the corresponding SPARQL endpoint using the documented anonymous credentials.

LLM angle none stated

Pitfalls & lessons The fetched page is an archived snapshot, and the provenance-rich dumps are much larger than the 315 MB vanilla HDT: listed variants range from 13.4 GB to 19 GB before zipped alternatives. Record both parts of the release identifier so the converter version and NELL iteration remain clear.

Verdict A provenance-conscious RDF publication of web-extracted general knowledge, best approached through the dump variant that matches the project’s storage and provenance needs.

Sources consulted

• https://web.archive.org/web/20241014060631/http://nell-ld.telecom-st-etienne.fr/