数据集:

cos_e

任务:

问答

子任务:

open-domain-qa

语言:

en

计算机处理:

monolingual

大小:

10K<n<100K

语言创建人:

crowdsourced

批注创建人:

crowdsourced

预印本库:

arxiv:1906.02361
英文

"cos_e" 数据集卡片

数据集摘要

Common Sense Explanations (CoS-E) 允许训练语言模型自动生成解释,这些解释可以在训练和推理中使用在新颖的常识自动生成解释(CAGE)框架中。

支持的任务和排行榜

More Information Needed

语言

More Information Needed

数据集结构

数据实例

v1.0
  • 下载的数据集文件大小:4.30 MB
  • 生成的数据集大小:2.34 MB
  • 总使用的磁盘空间:6.64 MB

"train" 的一个示例如下所示。

{
    "abstractive_explanation": "this is open-ended",
    "answer": "b",
    "choices": ["a", "b", "c"],
    "extractive_explanation": "this is selected train",
    "id": "42",
    "question": "question goes here."
}
v1.11
  • 下载的数据集文件大小:6.53 MB
  • 生成的数据集大小:3.05 MB
  • 总使用的磁盘空间:9.58 MB

"train" 的一个示例如下所示。

{
    "abstractive_explanation": "this is open-ended",
    "answer": "b",
    "choices": ["a", "b", "c"],
    "extractive_explanation": "this is selected train",
    "id": "42",
    "question": "question goes here."
}

数据字段

所有拆分的数据字段都相同。

v1.0
  • id:字符串特征。
  • question:字符串特征。
  • choices:字符串特征列表。
  • answer:字符串特征。
  • abstractive_explanation:字符串特征。
  • extractive_explanation:字符串特征。
v1.11
  • id:字符串特征。
  • question:字符串特征。
  • choices:字符串特征列表。
  • answer:字符串特征。
  • abstractive_explanation:字符串特征。
  • extractive_explanation:字符串特征。

数据拆分

name train validation
v1.0 7610 950
v1.11 9741 1221

数据集创建

策划理由

More Information Needed

源数据

初始数据收集和规范化

More Information Needed

谁是源语言的生产者?

More Information Needed

标注

标注过程

More Information Needed

谁是标注者?

More Information Needed

个人隐私信息

More Information Needed

使用数据的注意事项

数据集的社会影响

More Information Needed

偏见讨论

More Information Needed

其他已知限制

More Information Needed

其他信息

数据集策划者

More Information Needed

许可信息

未知。

引用信息

@inproceedings{rajani2019explain,
     title = "Explain Yourself! Leveraging Language models for Commonsense Reasoning",
    author = "Rajani, Nazneen Fatema  and
      McCann, Bryan  and
      Xiong, Caiming  and
      Socher, Richard",
      year="2019",
    booktitle = "Proceedings of the 2019 Conference of the Association for Computational Linguistics (ACL2019)",
    url ="https://arxiv.org/abs/1906.02361"
}

贡献

感谢 @lewtun @thomwolf @mariamabarham @patrickvonplaten @albertvillanova @lhoestq 添加此数据集。