where can i find the whole Train-Doc?

Train-Doc: For the remaining 4, 268 documents, we employ an automated method developed in our prior
work (Bhattacharya et al., 2019), which predicts the functional role for each sentence in Indian Supreme
Court judgment documents, and extract those sentences that are predicted as facts. This method achieved
an F1-score of 0.84 on identifying facts, so we consider it reliable for our purpose.