The SciREX-train.ipynb jupyter notebook extracts 1,000 sentences from the SciREX data set and creates our own annotation file. This file then needs to be labeled with used/proposed classification.
The Saliency Correlation notebook contains all code for calculating the correlation between salient entities and entities that we marked as used.