keepset

WN18RR

Heterogeneous knowledge graph derived from WordNet lexical database. Contains words and their semantic relationships (hypernyms, synonyms, etc.) with 11 relation types. Created to fix test set leakage in the original WN18 dataset by removing inverse relation redundancy. Standard benchmark for knowledge graph completion.

npx keepset fetch wn18rr

Nodes
40,943
Edges
93,003
Size
1.2 MiB

Fetch and preview this dataset on this page, or download the result as JSON. The command line fetches it without a browser.

Overview

Idwn18rr
Categorygraph/v1
Kindgraph
DomainKnowledge graph
Licencenot recorded by the source
Cataloguekeepset 1.33.0

Sources and provenance

#URLStatusRetrievedRead by recipePinned snapshotSHA-256
0https://web.informatik.uni-mannheim.de/pi1/kge-datasets/wnrr.tar.gzlive2026-03-03yes
1https://github.com/pykeen/pykeen#datasetslive2026-03-03

Citations

Recipe

Source#0
Archivetar.gz
Part trainwnrr/train.txt
Part validwnrr/valid.txt
Part testwnrr/test.txt
Parser format"kg-triples"
Parser encoding"utf-8"
Parser delimiter"tab"
Parser order"head-relation-tail"
Parser header"none"
Parser quotedfalse

Crosswalk

keepset/relations-split to graph/v1, as a crossmap crosswalk.

RuleFromToDirection
movemetametaboth
movenodes[].idnodes[].idboth
moveedges[].sourceedges[].sourceboth
moveedges[].targetedges[].targetboth
moveedges[].typeedges[].typeboth
moveedges[].splitedges[].splitboth

Expected counts

Edges93,003
Nodes40,943

Properties

Components13
Connectedno
Cyclicyes
Density0.0006859
Directedyes
Download bytes1.2 MiB
Edge types11
Heterogeneousyes
Multigraphyes
Node types1
Raw bytes0 B
Self loopsyes
Weightedno