This repository contains a script from the [DiscoDisco system](https://github.com/gucorpling/DisCoDisCo): features.py (originally found [here](https://github.com/gucorpling/DisCoDisCo/blob/master/gucorpling_models/rel/features.py)), with slight modifications. This script requires all files of the DISRPT 2021 data (not just .rels) and is **essential** to run the following experiments.
For each DISRPT 2021 corpus, please run the command:
```
python features.py \
--disrpt_data [PATH OF DISRPT 2021 DATA] \
--corpus [NAME OF CORPUS, e.g. eng.pdtb.pdtb] \
--save_dir [NEW FOLDER TO SAVE NESTED DATASETS]
```
This script will create folders and files with the same structure as DISRPT data, but with additional feature columns in .rels files.