This repository is the official implementation of Understanding and Improving Knowledge Graph Embedding for Entity Alignment, ICML 2022.
Please see the paper A Benchmarking Study of Embedding-based Entity Alignment for Knowledge Graphs (VLDB 2020) for dataset details.
This project is based on OpenEA. We did not add additional packages compared with the original OpenEA project.
We provide an example of jupyter notebook.
conda create -n openea python=3.6
conda activate openea
conda install tensorflow-gpu==1.8
conda install -c conda-forge graph-tool==2.29
conda install -c conda-forge python-igraph
pip install -r requirement.txt
pip install -e .
python main_from_args.py ./args/sea_args_15K.json D_W_15K_V1 721_5fold/1/
./src/openea/approaches/neural_ontology.py
We slightly modified the source code of the baselines to inject neural ontology into them:
BootEA: ./src/openea/approaches/bootea.py
SEA: ./src/openea/approaches/sea.py
RSN: ./src/openea/approaches/rsn4ea.py
RDGCN: ./src/openea/approaches/rdgcn.py
We added NeoEA hyper-parameters to the original setting files:
BootEA: ./run/args/bootea_args_15K.json
SEA: ./run/args/sea_args_15K.json
RSN: ./run/args/rsn4ea_args_15K.json
RDGCN: ./run/args/rdgcn_args_15K.json