BEADs

该数据集名为BEADs，旨在支持广泛的自然语言处理任务，包括文本分类、标记分类、偏见量化以及良性语言生成。它包含了大约370万条记录，这些记录集中关注偏见的各个方面。此外，该数据集还包括一个由专家验证的金标签子集，并采用了如主动学习等标注完善的方法。规模上，该数据集拥有大约370万条记录，适用于文本分类、标记分类、偏见量化和良性语言生成等多种任务。

This dataset, named BEADs, is designed to support a wide range of natural language processing (NLP) tasks, including text classification, token classification, bias quantification, and benign language generation. It contains approximately 3.7 million records that focus on various dimensions of bias. Additionally, the dataset includes a subset of expert-validated gold standard labels, and employs annotation refinement methods such as active learning. With a total of around 3.7 million records, it is suitable for the aforementioned diverse NLP tasks.