Decompressing Knowledge Graph Representations for Link Prediction

11 Nov 2019Xiang KongXianyang ChenEduard Hovy

This paper studies the problem of predicting missing relationships between entities in knowledge graphs through learning their representations. Currently, the majority of existing link prediction models employ simple but intuitive scoring functions and relatively small embedding size so that they could be applied to large-scale knowledge graphs... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Link Prediction FB15k-237 RESCAL + Decom MRR 0.354 # 10
Link Prediction FB15k-237 RESCAL + Decom [email protected] 0.536 # 8
Link Prediction FB15k-237 RESCAL + Decom [email protected] 0.388 # 6
Link Prediction FB15k-237 RESCAL + Decom [email protected] 0.261 # 7
Link Prediction WN18RR RESCAL + Decom MRR 0.457 # 17
Link Prediction WN18RR RESCAL + Decom [email protected] 0.515 # 21
Link Prediction WN18RR RESCAL + Decom [email protected] 0.469 # 13
Link Prediction WN18RR RESCAL + Decom [email protected] 0.427 # 13