NMN

Source code and datasets for ACL 2020 paper: Neighborhood Matching Network for Entity Alignment.

Datasets

Please first download the datasets here and extract them into data/ directory.

Initial datasets DBP15K and DWY100K are from JAPE and BootEA.

Take the dataset DBP15K (ZH-EN) as an example, the folder “zh_en” contains:

• 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 (DBP_ZH);
• triples_1_s: remaining relation triples encoded by ids in source KG (S-DBP_ZH);
• triples_2: relation triples encoded by ids in target KG (DBP_EN);
• triples_2_s: remaining relation triples encoded by ids in target KG (S-DBP_EN);
• vectorList.json: the input entity feature matrix initialized by word vectors;

Environment

• Python>=3.5
• Tensorflow>=1.8.0
• Scipy
• Numpy

Due to the limited graphics memory of GPU, we ran our codes using CPUs (40 Intel(R) Xeon(R) CPU E5-2640 v4 @ 2.40GHz).

Running

For example, to run NMN on DBP15K (ZH-EN), use the following script:

python3 main.py --dataset DBP15k --lang zh_en

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 wyting@pku.edu.cn.

Citation

If you use this model or code, please cite it as follows:

Yuting Wu, Xiao Liu, Yansong Feng, Zheng Wang and Dongyan Zhao. Neighborhood Matching Network for Entity Alignment. In: ACL 2020.