These are the splits I generated for my active learing blog. Each example in the training sets have five total annotations, either in `X_labeled_full` or `extra_annotations.npy`.

Files:
- `test.csv`: test split of high-quality examples
- `X_labeled_full`: starting set of labeled examples w/ 2 annotations each
- `X_unlabeled.csv`: collection of unlabeled examples
- `extra_annotations.npy`: extra annotations for training examples.