Large-scale heterogeneous knowledge graph from Wikidata 2021 dump. Contains ~5M entities connected by 822 relation types with ~21M triples. Each entity paired with Wikipedia article text for joint structured-unstructured modeling. Supports transductive and inductive link prediction settings.
npx keepset fetch wikidata5m
Nodes
4,594,485
Edges
20,624,575
Size
160.5 MiB
This entry has no fetchable recipe yet: no download is in a form keepset reads yet (kge-triples-tar: the triples are more edges than one fetch holds: a whole fetch at Node's default heap aborts with an allocation failure, because the crosswalk step copies every edge (ExaDev/crossmap#57))
Xiaozhi Wang, Tianyu Gao, Zhaochen Song, Zenan Li, Bing Qin, Tie-Yan Liu, Zhiyuan Liu, Jie Tang (2021). KEPLER: A Unified Model for Knowledge Embedding and Pre-trained Language Representation. Transactions of the Association for Computational Linguistics.
Recipe
Status
not fetchable: no download is in a form keepset reads yet (kge-triples-tar: the triples are more edges than one fetch holds: a whole fetch at Node's default heap aborts with an allocation failure, because the crosswalk step copies every edge (ExaDev/crossmap#57))
Crosswalk
keepset/topology to graph/v1, as a crossmap crosswalk.