Convolutional 2D Knowledge Graph Embeddings

5 Jul 2017Tim DettmersPasquale MinerviniPontus StenetorpSebastian Riedel

Link prediction for knowledge graphs is the task of predicting missing relationships between entities. Previous work on link prediction has focused on shallow, fast models which can scale to large knowledge graphs... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Link Prediction FB15k Inverse Model MR 2501 # 1
Link Prediction FB15k Inverse Model MRR 0.660 # 8
Link Prediction FB15k Inverse Model [email protected] 0.660 # 8
Link Prediction FB15k Inverse Model [email protected] 0.659 # 7
Link Prediction FB15k Inverse Model [email protected] 0.658 # 6
Link Prediction FB15k-237 ConvE MRR 0.325 # 6
Link Prediction FB15k-237 ConvE [email protected] 0.501 # 8
Link Prediction FB15k-237 ConvE [email protected] 0.356 # 5
Link Prediction FB15k-237 ConvE [email protected] 0.237 # 6
Link Prediction WN18 Inverse Model MRR 0.963 # 1
Link Prediction WN18 Inverse Model [email protected] 0.964 # 1
Link Prediction WN18 Inverse Model [email protected] 0.964 # 1
Link Prediction WN18 Inverse Model [email protected] 0.953 # 1
Link Prediction WN18 Inverse Model MR 740 # 2
Link Prediction WN18RR ConvE MRR 0.430 # 12
Link Prediction WN18RR ConvE [email protected] 0.520 # 11
Link Prediction WN18RR ConvE [email protected] 0.440 # 8
Link Prediction WN18RR ConvE [email protected] 0.400 # 10
Link Prediction YAGO3-10 ConvE MRR 0.44 # 4
Link Prediction YAGO3-10 ConvE [email protected] 0.62 # 4