The implementation of GAIL-TPath is based on the code released by https://github.com/DMKE-Lab/TPath
To install the various python dependences (including tensorflow)
pip install -r requirements.txt
The preprocessed NELL-995 dataset can be downloaded here and should be located in the datasets directory.
The hyperparameter configs for each experiments are in the configs directory. To start a particular experiment, just do
sh run.sh configs/${dataset}.sh
where the ${dataset}.sh is the name of the config file. For example,
sh run.sh configs/ICEWS14s.sh
We are also releasing pre-trained GAIL-TPath models for testing. They can be downloaded here and should be located in the output directory. To load the model, set the load_model to 1 in the config file (default value 0) and model_load_dir to point to the saved_model.
To start a particular experiment, just do
sh run.sh test_configs/${dataset}.sh
where the ${dataset}.sh is the name of the config file. For example,
sh run.sh test_configs/ICEWS14s.sh
To start the whole testing experiment, just do
python test_nell.py
The metrics used for evaluation are Hits@{1,3}, MAP and MRR. Along with this, the code also outputs the answers GAIL-TPath reached in a file.
If you use this code, please cite our paper