Paper: https://arxiv.org/abs/2203.09280
Due to licensing we are not allowed to distribute the datasets bbc-db, imdb-tmdb, tmdb-tvdb, imdb-tvdb, restaurants. To run the experiments, please download bbc-db, imdb, tmdb, tvdb and restaurants.
You should :
- Install Anaconda 4.11.0
- Download wiki-news-300d-1M.vec.zip from https://fasttext.cc/docs/en/english-vectors.html
- Unzip wiki-news-300d-1M.vec.zip
- Copy wiki-news-300d-1M.vec to OpenEA/datasets
cd AttrE
conda env create --file install/AttrE.yml --name AttrE_env
conda activate AttrE_env
cd "dataset_name" (e.g., restaurants)
python KBA.py
cd RREA_versions
conda env create --file install/RREA.yml --name RREA_env
conda activate RREA_env
python RREA.py
cd OpenEA
conda env create --file install/openea.yml --name OpenEA_env
conda activate OpenEA_env
pip install -e .
cd run
python main_from_args.py "predefined_arguments" "dataset_name" "split"
For example, for MTransE on tmdb-tvdb for the first split, please execute the following:
python main_from_args.py ./args/mtranse_args_15K.json D_W_15K_V1 721_5fold/1/
cd analysis
conda env create --file install/analysis.yml --name analysis_env
conda activate analysis_env
For Nemenyi Diagrams
python nemenyi_diagrams.py
For Correlations
python correlation_analysis.py
For Time to Accuracy
python time_to_accuracy.py
For Trade Offs
python trade_offs.py
please download descriptions_pickles.zip from https://www.dropbox.com/sh/7r2y1x8wx9y921e/AAAh3DabBguCzxk8OlZnL-Mza?dl=0
uzip descriptions_pickles.zip
copy to ./statistics/
cd statistics
conda env create --file install/statistics.yml --name statistics_env
conda activate statistics_env
python statistics.py