TuckER: Tensor Factorization for Knowledge Graph Completion

Knowledge graphs are structured representations of real world facts. However, they typically contain only a small subset of all possible facts. Link prediction is a task of inferring missing facts based on existing ones. We propose TuckER, a relatively straightforward but powerful linear model based on Tucker decomposition of the binary tensor representation of knowledge graph triples. TuckER outperforms previous state-of-the-art models across standard link prediction datasets, acting as a strong baseline for more elaborate models. We show that TuckER is a fully expressive model, derive sufficient bounds on its embedding dimensionalities and demonstrate that several previously introduced linear models can be viewed as special cases of TuckER.

PDF Abstract IJCNLP 2019 PDF IJCNLP 2019 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Link Prediction FB15k TuckER MRR 0.795 # 10
Hits@10 0.892 # 8
Hits@3 0.833 # 5
Hits@1 0.741 # 6
Link Prediction FB15k-237 TuckER MRR 0.358 # 19
Hits@10 0.544 # 19
Hits@3 0.394 # 14
Hits@1 0.266 # 15
Link Prediction WN18 TuckER MRR 0.953 # 2
Hits@10 0.958 # 10
Hits@3 0.955 # 2
Hits@1 0.949 # 2
Link Prediction WN18RR TuckER MRR 0.470 # 39
Hits@10 0.526 # 52
Hits@3 0.482 # 33
Hits@1 0.443 # 21

Methods