DualMatch

Source code for "Unsupervised Entity Alignment for Temporal Knowledge Graphs". The ACM Web Conference 2023.

A 3-minute short video briefly describing our method can be found Here.

Dataset

We thank TEA-GNN for providing the datasets.

-ICEWS05-15

-YAGO-WIKI50K

-YAGO-WIKI20K

ent_ids_1: ids for entities in source KG;
ent_ids_2: ids for entities in target KG;
ref_ent_ids: entity links encoded by ids;
triples_1: relation triples encoded by ids in source KG;
triples_2: relation triples encoded by ids in target KG;
rel_ids_1: ids for relations in source KG;
rel_ids_2: ids for relations in target KG;
sup_pairs + ref_pairs: entity alignments

Run

To perform EA on ICEWS05-15 in unsupervised manner:

python main.py --ds 0 --unsup

To perform EA on YAGO-WIKI50K in Less seed setting:

python main.py --ds 1 --train_ratio 20

To perform EA on YAGO-WIKI20K in Normal setting and evaluate separately:

python main.py --ds 2 --sep_eval

Note: Part of the code needs to be run on CPU. We plan to fix this issue in the future.

Acknowledgements

We refer to the code of Dual-AMN, TEA-GNN, and SEU. Thanks for their great contributions!

Code credit to Xiaoze Liu and Junyang Wu.

Citation

If you find this work useful, please cite

    @inproceedings{DBLP:conf/www/LiuW00G23,
      author       = {Xiaoze Liu and
                      Junyang Wu and
                      Tianyi Li and
                      Lu Chen and
                      Yunjun Gao},
      title        = {Unsupervised Entity Alignment for Temporal Knowledge Graphs},
      booktitle    = {Proceedings of the {ACM} Web Conference 2023, {WWW} 2023, Austin,
                      TX, USA, 30 April 2023 - 4 May 2023},
      pages        = {2528--2538},
      publisher    = {{ACM}},
      year         = {2023},
      url          = {https://doi.org/10.1145/3543507.3583381},
      doi          = {10.1145/3543507.3583381},
    }