UNcommonsense

# Dataset Card for UNcommonsense

## Dataset Description

- **Paper:** https://arxiv.org/abs/2311.08469
- **Point of Contact:** [Wenting Zhao](mailto:[e-mail])

### Dataset Summary

UNcommonsense is an abductive reasoning dataset. Unlike [aNLG](https://arxiv.org/abs/1908.05739), we focus on explaining unusual, unexpected, and unlikely situations. UNcommonsense is an English-language corpus consisting of 20k unique contexts paired with explicitly uncommon outcomes. Given these contexts and uncommon outcomes, we crowdsource 41k abductive explanations, which provide a plausible explanation of how an uncommon outcome could have arisen, given an input context.

### Data Fields

- `context` (string): Several sentences describing a context.
- `outcome` (string): An unexpected outcome from the context.
- `human_explanations` (list of strings): A list of human-authored explanations that make the unexpected outcome likely given the context.
- `gpt4_explanations` (list of strings): A list of GPT-4 generated explanations that make the unexpected outcome likely given the context.
- `enhanced_explanations` (list of strings): A list of GPT-4 enhanced human-authored explanations that make the unexpected outcome likely given the context.
- `source` (string): The source of the dataset from which we created the example.

### Citation Information

Please consider citing [our paper](https://arxiv.org/pdf/2311.08469.pdf) if you find this dataset useful:
```
@article{zhao2023uncommonsense,
title={UNcommonsense Reasoning: Abductive Reasoning about Uncommon Situations},
author={Zhao, Wenting and Chiu, Justin T and Hwang, Jena D and Brahman, Faeze and Hessel, Jack and Choudhury, Sanjiban and Choi, Yejin and Li, Xiang Lorraine and Suhr, Alane},
journal={arXiv preprint arXiv:2311.08469},
year={2023}
}
```

# UNcommonsense 数据集卡片

## 数据集描述
- **论文：** https://arxiv.org/abs/2311.08469
- **联系人：** [赵文婷](mailto:[e-mail])

### 数据集概览
UNcommonsense是一个溯因推理（abductive reasoning）数据集。不同于aNLG（https://arxiv.org/abs/1908.05739），本数据集聚焦于对反常、意外且不合常理的场景进行解释。UNcommonsense为英语语料库，包含2万个独特场景，并搭配明确的非典型结果。基于这些场景与非典型结果，我们通过众包构建了4.1万个溯因解释，用于阐明在给定输入场景的前提下，非典型结果如何能够合理产生。

### 数据字段
- `context`（字符串类型）：用于描述某一场景的若干语句。
- `outcome`（字符串类型）：由该场景衍生出的意外结果。
- `human_explanations`（字符串列表）：由人类撰写的解释集合，用于说明在给定场景下该意外结果具备合理性的缘由。
- `gpt4_explanations`（字符串列表）：由GPT-4生成的解释集合，用于说明在给定场景下该意外结果具备合理性的缘由。
- `enhanced_explanations`（字符串列表）：经GPT-4增强的人类撰写解释集合，用于说明在给定场景下该意外结果具备合理性的缘由。
- `source`（字符串类型）：构建该数据集示例的原始来源。

### 引用信息
若您认为本数据集对研究有所助益，请引用如下论文：

@article{zhao2023uncommonsense,
title={UNcommonsense Reasoning: Abductive Reasoning about Uncommon Situations},
author={Zhao, Wenting and Chiu, Justin T and Hwang, Jena D and Brahman, Faeze and Hessel, Jack and Choudhury, Sanjiban and Choi, Yejin and Li, Xiang Lorraine and Suhr, Alane},
journal={arXiv preprint arXiv:2311.08469},
year={2023}
}