Standard Staticcal Benchmark Datasets for Knowledge Graph Completion Github repository.
This repository contains dataset for Knowledge Graph Completion task include: link prediction (enity prediction, relation prediction). We upload common datasets like FB15K, FB15K-237, WN18, WN18RR, YAGO3-10. For each dataset, we extract 1-1, 1-N, N-1, N-N relation to text files, relation properties in dataset include symmetric, antisymmetric, reflexive, irreflexive, transitive, partial equivalence, equivalence, order and preorder type.
Contributed by: Nhut-Nam Le
FB15k-237 is upgraded version of FB15k. Inverse relations are deleted in order to prevent direct inference of test triples, model can't predict easily. In FB15K-237, we have some relation types like symmetric, antisymmetric and composite
WN18RR is a subset of WN18. Seven inverse relations are deleted similar to FB15k-237. It describe lexical and semantic hierarchies between concepts, which is mainly concerned with symmetry, anti-symmetry.
YAGO3-10 is a subset of YAGO3, mainly concerned with symmetry, anti-symmetry. It has 123,182 entities and 37 relations, and most of the triples describe attributes of persons such as citizenship, gender, and profession.
| Entities | Relations | Train | Valid | Test | |
|---|---|---|---|---|---|
| FB15K [1] | 14951 | 1345 | 483142 | 50000 | 50971 |
| WN18 [1] | 40943 | 18 | 141442 | 5000 | 5000 |
| FB15K-237 [2] | 14541 | 237 | 272115 | 17535 | 20466 |
| WN18RR [3] | 40943 | 11 | 86835 | 3034 | 3134 |
| YAGO3-10 [4] | 123182 | 37 | 1079040 | 5000 | 5000 |
| Kinship [5] | 104 | 25 | 8544 | 1068 | 1074 |
| UMLS | 135 | 46 | 5216 | 652 | 661 |
[1] Bordes, Antoine, Nicolas Usunier, Alberto García-Durán, Jason Weston and Oksana Yakhnenko. “Translating Embeddings for Modeling Multi-relational Data.” NIPS (2013).
[2] Toutanova, Kristina and Danqi Chen. “Observed versus latent features for knowledge base and text inference.” (2015).
[3] Dettmers, Tim, Pasquale Minervini, Pontus Stenetorp and Sebastian Riedel. “Convolutional 2D Knowledge Graph Embeddings.” AAAI (2018).
[4] Suchanek, Fabian M., Gjergji Kasneci and Gerhard Weikum. “Yago: a core of semantic knowledge.” WWW '07 (2007).
[5] Lin, Xi Victoria, Richard Socher and Caiming Xiong. “Multi-Hop Knowledge Graph Reasoning with Reward Shaping.” EMNLP (2018).
[6] Kok, Stanley and Pedro M. Domingos. “Statistical predicate invention.” ICML '07 (2007).
Add more dataset
research.lpca: project analyzes the results of various models for Link Prediction on Knowledge Graphs using Knowledge Graph Embeddings
RotatE: Knowledge Graph Embedding by Relational Rotation in Complex Space
OpenKE-PyTorch: An Open-Source Package for Knowledge Embedding (KE)