Scalable and interpretable rule-based link prediction for large heterogeneous knowledge graphs

Neural embedding-based machine learning models have shown promise for predicting novel links in biomedical knowledge graphs. Unfortunately, their practical utility is diminished by their lack of interpretability... (read more)

PDF Abstract

Datasets


Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK BENCHMARK
Link Prediction FB15k-237 SAFRAN (white box, rule based) Hits@10 0.5465 # 13
Hits@3 0.4175 # 3
Hits@1 0.3013 # 5
Link Prediction OpenBioLink DistMult Hits@10 0.534 # 3
Hits@3 0.331 # 4
Hits@1 0.184 # 4
Link Prediction OpenBioLink ComplEx Hits@10 0.525 # 4
Hits@3 0.314 # 6
Hits@1 0.166 # 5
Link Prediction OpenBioLink TransE Hits@10 0.441 # 7
Hits@3 0.268 # 7
Hits@1 0.128 # 7
Link Prediction OpenBioLink RotatE Hits@10 0.522 # 5
Hits@3 0.315 # 5
Hits@1 0.156 # 6
Link Prediction OpenBioLink SAFRAN (white box, rule based) Hits@10 0.5110 # 6
Hits@3 0.3473 # 3
Hits@1 0.2232 # 3
Link Prediction OpenBioLink TransR Hits@10 0.592 # 2
Hits@3 0.451 # 2
Hits@1 0.369 # 2
Link Prediction OpenBioLink RESCAL Hits@10 0.615 # 1
Hits@3 0.479 # 1
Hits@1 0.407 # 1

Methods used in the Paper


METHOD TYPE
SAFRAN
Rule-based systems
Symbolic rule learning
Rule-based systems