Source code and datasets for 2022 paper: [Multi-Heterogeneous Neighborhood-Aware for Knowledge Graphs Alignment, IPM, 2022]
Please first download the datasets here and extract them into
datasets/directory.
Initial datasets WN31-15K and DBP-15K are from OpenEA and JAPE.
Initial datasets DWY100K is from BootEA and MultiKE.
Take the dataset EN_DE(V1) as an example, the folder "pre " contains:
- kg1_ent_dict: ids for entities in source KG;
- kg2_ent_dict: ids for entities in target KG;
- ref_ent_ids: entity links encoded by ids;
- rel_triples_id: relation triples encoded by ids;
- attr_triples_id: attribute triples encoded by ids;
- kgs_num: statistics of the number of entities, relations, attributes, and attribute values;
- value_embedding.out: the input entity feature matrix initialized by word vectors;
- entity_embedding.out: the input attribute value feature matrix initialized by word vectors;
- Python>=3.7
- pytorch>=1.7.0
- tensorboardX>=2.1.0
- Numpy
- json
To run MHNA model on WN31-15K and DBP-15K, use the following script:
python3 align_exc_15K.py
To run MHNA model DWY100K, use the following script:
python3 align_exc_DWY100K.py
Due to the instability of embedding-based methods, it is acceptable that the results fluctuate a little bit (±1%) when running code repeatedly.
If you have any difficulty or question in running code and reproducing expriment results, please email to cwswork@qq.com.
If you use this model or code, please cite it as follows:
*Weishan Cai, Yizhao Wang, Shun Mao, Jieyu Zhan and Yuncheng Jiang. Multi-heterogeneous neighborhood-aware for knowledge graphs alignment. Information Processing & Management, 2022, 59(1): 102790.