KGC

This repository contains the code and example data for the paper "Supervised Cardinality Estimation for Similarity Queries on Knowledge Graphs".

Additional artifacts can be found as follows:

• LUBM: https://swat.cse.lehigh.edu/projects/lubm/
• WIKI: https://github.com/MillenniumDB/WDBench
• YELP: https://www.yelp.com/dataset
• SWDF: hhttp://www.scholarlydata.org
• LMKG code can be found here

A subset of the training and testing data for the LUBM dataset is provided. To extract training data for the rest of the datasets, the corresponding neo4j knowledge graph must be built. Then the scripts in the data_extraction folder may be used.

The code was developed using Python 3.10.4, keras 2.7.0 and tensorflow 2.7.1.

To run the script for the star similarity query example, please issue the following command:

python kgc_exp.py star 5 lubm 1 100 128

For the chain similarity query:

python kgc_exp.py chain 2 lubm 1 100 128

For more information regarding the arguments of the script, please refer to the kgc_exp.py file. When the code is executed, the trained models and results are stored in the models and results folders respectively. After execution, the user may find the set of results under a subfolder named according to the used dataset.