Source code for the CAFE tool, which evaluates triples for KG Completion.
This repository contains the necessary code to generate feature vectors using our proposed context-aware features, as well as the datasets that we used in our paper. To run CAFE, install the dependencies listed in requirements.txt and run python main.py <dataset> <max-ctx>, where <dataset> is the name of the dataset to use (which should be in the datasets/ folder) and <max-ctx> is the maximum path length used to generate neighborhood subgraphs.
If you find CAFE useful, please consider citing it as:
@article{borrego2021CAFE,
author = {Borrego, Agust{\'i}n and Ayala, Daniel and Hern{\'a}ndez, Inma and Rivero, Carlos R. and Ruiz, David},
title = {{CAFE}: Knowledge graph completion using neighborhood-aware features},
journal = {Engineering Applications of Artificial Intelligence},
volume = {103},
pages = {104302},
year = {2021},
issn = {0952-1976},
doi = {10.1016/j.engappai.2021.104302},
url = {https://www.sciencedirect.com/science/article/pii/S0952197621001500}
}