XStoryCloze

Summary

• Introduction
• Dataset Structure
• Reference
• License
• Citation

Introduction

XStoryCloze consists of the professionally translated version of the English StoryCloze dataset (Spring 2016 version) to 10 non-English languages. This dataset is released by Meta AI.

Dataset Structure

Data Instances

• Size of downloaded dataset files: 2.03 MB
• Size of the generated dataset: 2.03 MB
• Total amount of disk used: 2.05 MB

An example of 'train' looks as follows.

{
    'answer_right_ending': 1,
    'input_sentence_1': 'Rick grew up in a troubled household.',
    'input_sentence_2': 'He never found good support in family, and turned to gangs.',
    'input_sentence_3': "It wasn't long before Rick got shot in a robbery.",
    'input_sentence_4': 'The incident caused him to turn a new leaf.',
    'sentence_quiz1': 'He is happy now.',
    'sentence_quiz2': 'He joined a gang.',
    'story_id': '138d5bfb-05cc-41e3-bf2c-fa85ebad14e2'
}

Data Fields

The data fields are the same among all splits.

• input_sentence_1: The first statement in the story.
• input_sentence_2: The second statement in the story.
• input_sentence_3: The third statement in the story.
• input_sentence_4: The forth statement in the story.
• sentence_quiz1: first possible continuation of the story.
• sentence_quiz2: second possible continuation of the story.
• answer_right_ending: correct possible ending, either 1 or 2.
• story_id: story id.

Data Splits

This dataset is intended to be used for evaluating the zero- and few-shot learning capabilities of multlingual language models. We split the data for each language into train and test (360 vs. 1510 examples, respectively). The released data files for different languages maintain a line-by-line alignment.

name	train	test
en	360	1510
ru	360	1510
zh	360	1510
es	360	1510
ar	360	1510
hi	360	1510
id	360	1510
te	360	1510
sw	360	1510
eu	360	1510
my	360	1510

Reference

We would like to acknowledge Xi Victoria Lin and Todor Mihaylov et al. for creating and maintaining the XStoryCloze dataset as a valuable resource for the computer vision and machine learning research community. For more information about the XStoryCloze dataset and its creator, please visit The XStoryCloze website.

License

The dataset has been released under the Creative Commons Attribution-ShareAlike 4.0 International License.

Citation

@article{DBLP:journals/corr/abs-2112-10668,
  author    = {Xi Victoria Lin and
               Todor Mihaylov and
               Mikel Artetxe and
               Tianlu Wang and
               Shuohui Chen and
               Daniel Simig and
               Myle Ott and
               Naman Goyal and
               Shruti Bhosale and
               Jingfei Du and
               Ramakanth Pasunuru and
               Sam Shleifer and
               Punit Singh Koura and
               Vishrav Chaudhary and
               Brian O'Horo and
               Jeff Wang and
               Luke Zettlemoyer and
               Zornitsa Kozareva and
               Mona T. Diab and
               Veselin Stoyanov and
               Xian Li},
  title     = {Few-shot Learning with Multilingual Language Models},
  journal   = {CoRR},
  volume    = {abs/2112.10668},
  year      = {2021},
  url       = {https://arxiv.org/abs/2112.10668},
  eprinttype = {arXiv},
  eprint    = {2112.10668},
  timestamp = {Tue, 04 Jan 2022 15:59:27 +0100},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2112-10668.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}