We've identified articles which were retracted. We need to get a sample of non-retracted articles and create a train and test set.