Structural Bias in Knowledge Graphs for the Entity Alignment Task

The source code of the paper:

N. Fanourakis, V. Efthymiou, V. Christophides, D. Kotzinos, E. Pitoura, K. Stefanidis. Structural Bias in Knowledge Graphs for the Entity Alignment Task. ESWC 2023 [paper]

Getting Started

Creating conda environment for RREA, sampling and analysis:

conda env create --file install/env.yml -n dev_env

Creating conda environment for RDGCN and MultiKE:

cd RDGCN
conda env create --file install/OpenEA.yml -n openea_env

Creating conda environment for PARIS:

cd PARIS
conda env create --file install/entity_match.yml -n entity_match

Execution Examples

For sampling:

python main.py start_sampling "configuration_name"

For analysis:

python main.py start_analysis sampled "configuration_name"

For running RREA:

python RREA.py original original (for original datasets)

or

python RREA.py sampled "configuration_name" (for sampled datasets)

For running RDGCN you should first:

1. Download wiki-news-300d-1M.vec.zip from https://fasttext.cc/docs/en/english-vectors.html
2. Unzip wiki-news-300d-1M.vec.zip
3. Copy wiki-news-300d-1M.vec to /datasets
4. cd RDGCN and pip install -e .

cd run
python main_from_args.py args/original_rdgcn_args_15K.json (for original datasets)

or

cd run
python main_from_args.py args/sampled_rdgcn_args_15K.json (for sampled datasets)

For running MultiKE:

cd run
python main_from_args_wo_attr.py args/original_multike_args_15K.json (for original datasets)

or

cd run
python main_from_args_wo_attr.py args/sampled_multike_args_15K.json (for sampled datasets)

For running PARIS:

cd src/experiments
mkdir results
python3 -u ../run_experiment.py \
        --method PARIS\
        --root_dataset "root of dataset folder"\
        --dataset "configuration_name"\
        --dataset_division 721_5fold\
        --out_folder ./results\
        --use_func > test.log

or

cd src/experiments
mkdir results
python3 -u ../run_experiment.py \
        --method PARIS\
        --root_dataset "root of dataset folder"\
        --dataset original\
        --dataset_division 721_5fold\
        --out_folder ./results\
        --use_func > test.log