# ROCStories 2018 Dataset

This is the HuggingFace version of ROCStories Dataset.

We take a few changes.

1. Column names are renamed to match in each split, for example in `test` we have originally `InputSentenceX`, that was renamed do `sentenceX` to match the `train` split.

2. Columns missing in different splits were add to the other split with empty values. For example `train` split contains `storytitle` column. This column as added in `test` and `validation` splits. 

3. There was three columns in `test`/`validation` splits. `RandomFifthSentenceQuiz[1|2]` and `AnswerRightEnding`. The first ones where renamed to `sentence5` if they are the correct answer (based on the `AnserRightEnding`) and the other was renamed to `sentenceE` (Error) if its not the right answer.

WARNING: The test set of 2018 dataset do not contains the column `AnswerRightEnding`, therefore, for sake of completeness we took `RandomFifthSentenceQuiz1` as the correct one and the `RandomFifthSentenceQuiz1` as the wrong one. If you know where to find the correct answers, please let me know and I will fix the dataset.

Most of these changes were taken to avoid a limitation in DatasetDicts in HuggingFace, since they all need to have the same columns.

Regarding the ROCStories these are the original papers

"Tackling The Story Ending Biases in The Story Cloze Test". Rishi Sharma, James Allen, Omid Bakhshandeh, Nasrin Mostafazadeh. In Proceedings of the 2018 Conference of the Association for Computational Linguistics (ACL), 2018