aieng-lab/biasneutral

Dataset Name数据集是BookCorpus数据集的过滤版本，去除了与性别、种族和宗教相关的偏见词汇，只包含中性词汇。该数据集适用于训练和评估语言模型，特别是在无偏见性别、种族和宗教方面的模型性能。数据集通过特定的过滤条件从BookCorpus中生成，包括字符长度限制、不包含特定名称、性别代词、性别名词和偏见属性词。

Dataset Name is a filtered version of the BookCorpus dataset, containing only bias-neutral words related to gender, race, and religion. It is suitable for training and evaluating language models, particularly in terms of model performance in unbiased gender, race, and religion. The dataset is generated from BookCorpus through specific filtering criteria, including character length restrictions, exclusion of specific names, gender pronouns, gender nouns, and bias attribute words.