Heterogeneous knowledge graph from Freebase with 237 relation types. A subset of FB15k created by removing inverse relations to prevent test leakage. Contains entities from multiple domains (people, locations, music, film) with directed relation edges. Standard benchmark for knowledge graph completion and link prediction.
npx keepset fetch fb15k-237
Nodes
14,541
Edges
310,116
Size
3.7 MiB
Fetch and preview this dataset on this page, or download the result as JSON. The command line fetches it without a browser.
Kristina Toutanova, Danqi Chen, Patrick Pantel, Hoifung Poon, Pallavi Choudhury, Michael Gamon (2015). Representing Text for Joint Embedding of Text and Knowledge Bases. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing (EMNLP).
Recipe
Source
#0
Archive
tar.gz
Part train
fb15k-237/train.txt
Part valid
fb15k-237/valid.txt
Part test
fb15k-237/test.txt
Parser format
"kg-triples"
Parser encoding
"utf-8"
Parser delimiter
"tab"
Parser order
"head-relation-tail"
Parser header
"none"
Parser quoted
false
Crosswalk
keepset/relations-split to graph/v1, as a crossmap crosswalk.