All commonsense causal data is in COPES.json. The data is recorded in the format of json records with three fileds:
- story: the story of 5 sentences sampled from RocStories.
- cause_idx: the events that have a causal relation with the last event (indices start from 0).
- res_idx: the result, just the last event.
The split_idx.json provide the indices for validation and testing data, indexing from 0.