README.md

metadata

language:
  - en
  - fr
  - ja
  - pt
  - ru
  - zh
license: cc-by-4.0
dataset_info:
  - config_name: en
    features:
      - name: sentence
        dtype: string
      - name: option1
        dtype: string
      - name: option2
        dtype: string
      - name: answer
        dtype: string
    splits:
      - name: test
        num_bytes: 282033
        num_examples: 2325
    download_size: 135735
    dataset_size: 282033
  - config_name: fr
    features:
      - name: sentence
        dtype: string
      - name: option1
        dtype: string
      - name: option2
        dtype: string
      - name: answer
        dtype: string
    splits:
      - name: test
        num_bytes: 10323
        num_examples: 83
    download_size: 7763
    dataset_size: 10323
  - config_name: jp
    features:
      - name: sentence
        dtype: string
      - name: option1
        dtype: string
      - name: option2
        dtype: string
      - name: answer
        dtype: string
    splits:
      - name: test
        num_bytes: 153848
        num_examples: 959
    download_size: 71460
    dataset_size: 153848
  - config_name: pt
    features:
      - name: sentence
        dtype: string
      - name: option1
        dtype: string
      - name: option2
        dtype: string
      - name: answer
        dtype: string
    splits:
      - name: test
        num_bytes: 33333
        num_examples: 263
    download_size: 17875
    dataset_size: 33333
  - config_name: ru
    features:
      - name: sentence
        dtype: string
      - name: option1
        dtype: string
      - name: option2
        dtype: string
      - name: answer
        dtype: string
    splits:
      - name: test
        num_bytes: 73406
        num_examples: 315
    download_size: 30788
    dataset_size: 73406
  - config_name: zh
    features:
      - name: sentence
        dtype: string
      - name: option1
        dtype: string
      - name: option2
        dtype: string
      - name: answer
        dtype: string
    splits:
      - name: test
        num_bytes: 105822
        num_examples: 504
    download_size: 82106
    dataset_size: 105822
configs:
  - config_name: en
    data_files:
      - split: test
        path: en/test-*
  - config_name: fr
    data_files:
      - split: test
        path: fr/test-*
  - config_name: jp
    data_files:
      - split: test
        path: jp/test-*
  - config_name: pt
    data_files:
      - split: test
        path: pt/test-*
  - config_name: ru
    data_files:
      - split: test
        path: ru/test-*
  - config_name: zh
    data_files:
      - split: test
        path: zh/test-*

XWinograd

Multilingual winograd schema challenge as used in Crosslingual Generalization through Multitask Finetuning.

Languages & Samples

• "en": 2325
• "fr": 83
• "jp": 959
• "pt": 263
• "ru": 315
• "zh": 504

Dataset creation

The Winograd schema challenges in this dataset combine winograd schemas from the XWinograd dataset introduced in Tikhonov et al and as it only contains 16 Chinese schemas, we add 488 Chinese schemas from clue/cluewsc2020.

If you only want the original xwinograd chinese schemas only, do:

load_dataset("Muennighoff/xwinograd", "zh")["test"][0][:16]

Additional Information

Citation Information

@misc{muennighoff2022crosslingual,
      title={Crosslingual Generalization through Multitask Finetuning}, 
      author={Niklas Muennighoff and Thomas Wang and Lintang Sutawika and Adam Roberts and Stella Biderman and Teven Le Scao and M Saiful Bari and Sheng Shen and Zheng-Xin Yong and Hailey Schoelkopf and Xiangru Tang and Dragomir Radev and Alham Fikri Aji and Khalid Almubarak and Samuel Albanie and Zaid Alyafeai and Albert Webson and Edward Raff and Colin Raffel},
      year={2022},
      eprint={2211.01786},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

@misc{tikhonov2021heads,
    title={It's All in the Heads: Using Attention Heads as a Baseline for Cross-Lingual Transfer in Commonsense Reasoning},
    author={Alexey Tikhonov and Max Ryabinin},
    year={2021},
    eprint={2106.12066},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

License

Like the original English winograd schema challenge, this dataset is licensed under CC BY 4.0. I.e. you can use it for commercial purposes etc. :)

Contributions

Thanks to Jordan Clive, @yongzx & @khalidalt for support on adding Chinese.