Text clustering applied to data augmentation in legal contexts