I wanted to use DuckDB using SQL to query Wikidata locally on my laptop. So I converted Wikidata's N-Triples dump (~8 billion rows) into Parquet files.
The dataset includes the full truth triplets (~60GB) and some pre-extracted datasets for YouTube channels, Letterboxd films, GitHub users and more.
The dataset includes the full truth triplets (~60GB) and some pre-extracted datasets for YouTube channels, Letterboxd films, GitHub users and more.
There's also a small website to explore random entities using DuckDB WASM: https://piebro.github.io/wikidata-extraction/random.html