A Benchmark Dataset for Learning to Intervene in Online Hate Speech
收藏资源简介:
为了鼓励对抗在线仇恨言论的策略,我们引入了一个新的生成式仇恨言论干预任务,并从Gab和Reddit收集了两个完全标记的数据集。与现有的仇恨言论数据集不同,我们的数据集保留了对话上下文并引入了人工编写的干预响应。由于我们的数据收集策略,我们数据集中的所有帖子都由Mechanical Turk工作者手动标记为仇恨或非仇恨言论,因此它们也可以用于仇恨言论检测任务。
To encourage strategies for combating online hate speech, we introduce a new generative hate speech intervention task and have collected two fully annotated datasets from Gab and Reddit. Unlike existing hate speech datasets, our datasets preserve conversational context and introduce human-authored intervention responses. Due to our data collection strategy, all posts in our datasets are manually labeled as hate or non-hate speech by Mechanical Turk workers, making them also suitable for hate speech detection tasks.
数据集概述
数据集名称
A Benchmark Dataset for Learning to Intervene in Online Hate Speech
数据集来源
- Gab
数据集特点
- 包含对话上下文
- 引入人工编写的干预响应
- 所有帖子由Mechanical Turk工作者手动标记为仇恨或非仇恨言论
数据集文件
gab.csvreddit.csv
数据结构
| 字段 | 描述 |
|---|---|
| id | 对话段落中帖子的ID |
| text | 对话段落中帖子的文本内容 |
| hate_speech_idx | 该对话中仇恨帖子的索引列表 |
| response | 人工编写的响应列表 |
数据集使用许可
- 根据Creative Commons Attribution-NonCommercial 4.0 International Public License授权




