PARADEHATE
收藏arXiv2025-06-02 更新2025-11-28 收录
下载链接：
https://hf-mirror.com/datasets/ScaDSAI/ParaDeHate
下载链接
链接失效反馈官方服务：
资源简介：
PARADEHATE是一个大规模的平行数据集，专门用于仇恨言论的解毒。该数据集包含超过8000对仇恨/非仇恨文本，由德国德累斯顿工业大学的研究团队创建。数据集的构建过程使用了大型语言模型（LLM），通过自动化的方式生成和验证文本，从而降低了数据创建的成本。该数据集可用于训练和评估解毒模型，以解决在线平台上日益增长的仇恨言论问题。
PARADEHATE is a large-scale parallel dataset specifically designed for hate speech detoxification. Comprising over 8,000 pairs of hate/non-hate texts, this dataset was developed by the research team from the Technical University of Dresden, Germany. The dataset was constructed using large language models (LLMs), which automatically generate and validate texts to reduce the costs of data creation. This dataset can be used to train and evaluate detoxification models to address the growing problem of hate speech on online platforms.
提供机构：
德国德累斯顿工业大学
创建时间：
2025-06-02
