fairnlp/holistic-bias

---
configs:
- config_name: noun_phrases
data_files:
- split: test
path: nouns.csv
- config_name: sentences
data_files:
- split: test
path: sentences.csv
license: cc-by-sa-4.0
language: en
---

# Usage

When downloading, specify which files you want to download and set the split to `train` (required by `datasets`).

```python
from datasets import load_dataset

nouns = load_dataset("fairnlp/holistic-bias", data_files=["nouns.csv"], split="train")
sentences = load_dataset("fairnlp/holistic-bias", data_files=["sentences.csv"], split="train")
```

# Dataset Card for Holistic Bias

This dataset contains the source data of the Holistic Bias dataset as described [by Smith et. al. (2022)](https://arxiv.org/abs/2205.09209). The dataset contains noun phrases and sentences used to measure the likelihood bias of various models. The original dataset is released on [GitHub](https://github.com/facebookresearch/ResponsibleNLP/tree/main/holistic_bias).

Disclaimer: this re-release of the dataset is not associated with the original authors. The dataset is released under the [CC-BY-SA-4.0](https://creativecommons.org/licenses/by-sa/4.0/) license.

## Dataset Details

The data is generated from the [official generation script](https://github.com/facebookresearch/ResponsibleNLP/blob/main/holistic_bias/generate_sentences.py). The data is the v1.0 data from the original paper.

For details on the methodology, please refer to the original paper. This dataset is contributed to Hugging
Face as part of the [FairNLP `fairscore` library](https://github.com/FairNLP/fairscore/).

### Dataset Sources

- **Paper:** https://arxiv.org/pdf/2205.09209.pdf

**BibTeX:**

```bibtex
@inproceedings{smith2022m,
title={“I’m sorry to hear that”: Finding New Biases in Language Models with a Holistic Descriptor Dataset},
author={Smith, Eric Michael and Hall, Melissa and Kambadur, Melanie and Presani, Eleonora and Williams, Adina},
booktitle={Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages={9180--9211},
year={2022}
}
```

原始信息汇总

数据集卡片摘要

数据集概述

• 名称: Holistic Bias
• 描述: 该数据集包含名词短语和句子，用于测量各种模型的偏见倾向。数据集来源于Smith等人的研究（2022年）。
• 许可: CC-BY-SA-4.0
• 语言: 英语

数据集详情

• 数据来源: 数据生成自官方生成脚本，版本为v1.0。
• 数据文件:
  • 名词短语: nouns.csv
  • 句子: sentences.csv

参考文献

• 论文: “I’m sorry to hear that”: Finding New Biases in Language Models with a Holistic Descriptor Dataset
• 作者: Smith, Eric Michael 等
• 会议: 2022 Conference on Empirical Methods in Natural Language Processing

搜集汇总

背景与挑战

背景概述

该数据集是一个用于检测语言模型偏见的文本数据集，包含大量名词短语和句子，适用于自然语言处理领域的偏见研究。数据集基于CC-BY-SA-4.0许可证发布，来源于相关学术论文的研究成果。