Is the sampled 120 BBC data used for evaluation in the paper stored in the github as well? There is a BBC data folder, but it has way more docs than 120 documents. I could not find the evaluation subset.