gokstad/toxic-nontoxic

这是一个包含Civil Comments平台评论的数据集，每个评论都有是否具有毒性的标注。数据集包含50k个训练示例，每个示例由10个标注者进行标注。当毒性评分大于等于0.5时，评论被标记为有毒。数据集中仅有约8%的评论被标记为有毒，显示出数据集的不平衡性。

This dataset includes comments from the Civil Comments platform, each with annotations indicating whether it is toxic or not. It contains 50k training examples, with each example annotated by 10 annotators. A comment is marked as toxic when the toxicity score is greater than or equal to 0.5. Only about 8% of the comments in the dataset are marked as toxic, indicating the imbalance of the dataset.