Ivana Balažević, Carl Allen, Timothy M. Hospedales
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.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Link Prediction | FB15k | Hits@1 | 0.741 | TuckER |
| Link Prediction | FB15k | Hits@10 | 0.892 | TuckER |
| Link Prediction | FB15k | Hits@3 | 0.833 | TuckER |
| Link Prediction | FB15k | MRR | 0.795 | TuckER |
| Link Prediction | WN18RR | Hits@1 | 0.443 | TuckER |
| Link Prediction | WN18RR | Hits@10 | 0.526 | TuckER |
| Link Prediction | WN18RR | Hits@3 | 0.482 | TuckER |
| Link Prediction | WN18RR | MRR | 0.47 | TuckER |
| Link Prediction | WN18 | Hits@1 | 0.949 | TuckER |
| Link Prediction | WN18 | Hits@10 | 0.958 | TuckER |
| Link Prediction | WN18 | Hits@3 | 0.955 | TuckER |
| Link Prediction | WN18 | MRR | 0.953 | TuckER |
| Link Prediction | FB15k-237 | Hits@1 | 0.266 | TuckER |
| Link Prediction | FB15k-237 | Hits@10 | 0.544 | TuckER |
| Link Prediction | FB15k-237 | Hits@3 | 0.394 | TuckER |
| Link Prediction | FB15k-237 | MRR | 0.358 | TuckER |