The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider removing the loading script and relying on automated data support (you can use convert_to_parquet from the datasets library). If this is not possible, please open a discussion for direct help.

Dataset Card for RiddleSense

Dataset Summary

Answering such a riddle-style question is a challenging cognitive process, in that it requires complex commonsense reasoning abilities, an understanding of figurative language, and counterfactual reasoning skills, which are all important abilities for advanced natural language understanding (NLU). However, there is currently no dedicated datasets aiming to test these abilities. Herein, we present RiddleSense, a new multiple-choice question answering task, which comes with the first large dataset (5.7k examples) for answering riddle-style commonsense questions. We systematically evaluate a wide range of models over the challenge, and point out that there is a large gap between the best-supervised model and human performance  suggesting intriguing future research in the direction of higher-order commonsense reasoning and linguistic creativity towards building advanced NLU systems.

Supported Tasks and Leaderboards

[Needs More Information]

Languages

English

Dataset Structure

Data Instances

An example of 'train' looks as follows.

{
    "answerKey": "E",
    "choices": {
        "label": ["A", "B", "C", "D", "E"],
        "text": ["throw", "bit", "gallow", "mouse", "hole"]
    },
    "question": "A man is incarcerated in prison, and as his punishment he has to carry a one tonne bag of sand backwards and forwards across a field the size of a football pitch.  What is the one thing he can put in it to make it lighter?"
}

Data Fields

Data Fields The data fields are the same among all splits.

default

  • answerKey: a string feature.
  • question: a string feature.
  • choices: a dictionary feature containing:
    • label: a string feature.
    • text: a string feature.

Data Splits

name train validation test
default 3510 1021 1184

Dataset Creation

Curation Rationale

[Needs More Information]

Source Data

Initial Data Collection and Normalization

[Needs More Information]

Who are the source language producers?

[Needs More Information]

Annotations

Annotation process

[Needs More Information]

Who are the annotators?

[Needs More Information]

Personal and Sensitive Information

[Needs More Information]

Considerations for Using the Data

Social Impact of Dataset

[Needs More Information]

Discussion of Biases

[Needs More Information]

Other Known Limitations

Dataset provided for research purposes only. Please check dataset license for additional information.

Additional Information

Dataset Curators

[Needs More Information]

Licensing Information

The copyright of RiddleSense dataset is consistent with the terms of use of the fan websites and the intellectual property and privacy rights of the original sources. All of our riddles and answers are from fan websites that can be accessed freely. The website owners state that you may print and download material from the sites solely for non-commercial use provided that we agree not to change or delete any copyright or proprietary notices from the materials. The dataset users must agree that they will only use the dataset for research purposes before they can access the both the riddles and our annotations. We do not vouch for the potential bias or fairness issue that might exist within the riddles. You do not have the right to redistribute them. Again, you must not use this dataset for any commercial purposes.

Citation Information

@InProceedings{lin-etal-2021-riddlesense,
title={RiddleSense: Reasoning about Riddle Questions Featuring Linguistic Creativity and Commonsense Knowledge},
author={Lin, Bill Yuchen and Wu, Ziyi and Yang, Yichi and Lee, Dong-Ho and Ren, Xiang},
journal={Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL-IJCNLP 2021): Findings},
year={2021}
}

Contributions

Thanks to @ziyiwu9494 for adding this dataset.

Downloads last month
192

Models trained or fine-tuned on INK-USC/riddle_sense