Enhanced Knowledge Graph Embedding by Jointly Learning Soft Rules and Facts.
Paper is published in MDPI Algorithms: download here.
We implement the embedding model of our algorithm SoLE based on the open resource code tensorflow-efe, where it trains the models by both training and validation sets while SoLE only uses training sets.
Thank the author of tensorflow-efe very much for sharing it.
- Python 3
- Tensorflow >= 1.2
- Hyperopt
- JBoss Drools 6.2.0
Datasets used in our paper are stored in the Datasets directory, including FB15K, DB100K and FB15K-sparse.
java -jar amie_plus.jar -maxad 3 -minpca 0.8 -minhc 0.8 ./Datasets/fb15k/train.txt
Rules extracted by AMIE+ are stored in the file rule_[confidence].txt which can be found in the corresponding dataset directory.
The project GenGroundings functions this module. After executing GroundAllRulesByRE.java, it will performs the reasoning and generates the groundings in the file groundings_[confidence].txt or groundings_oneTime_[confidence].txt.
The configurations of datasets can be set in config.py
Add hyperparameters dict and its identifier in model_param_space.py.
CUDA_VISIBLE_DEVICES=[gpu] python train.py -m [model_name] -d [data_name]
Train on the given hyperparameter setting and give the result for the test set.
If you find our work useful, please cite:
@article{zhang2019enhanced,
title={Enhanced Knowledge Graph Embedding by Jointly Learning Soft Rules and Facts},
author={Zhang, Jindou and Li, Jing},
journal={Algorithms},
volume={12},
number={12},
pages={265},
year={2019}}
MIT