Code needed to compute the features used by ResearchLink, condensed into a single file for convenience.
The hypotheses must be in the file data/hypotheses.csv with the following format: subject,predicate,object,label,scicheck_score. Then, execute run.py to generate a processed file with all features computed for each hypothesis, which can then be passed on to a classifier.
A number of precomputations are needed for different features. In the data/ folder, we provide all of them for the CSKG-600 dataset.