Trawling for Trolling

数据集'Trawling for Trolling'由印度国际信息技术学院创建，包含12,490个样本，分为5个类别：正常、亵渎、挑衅、贬低和仇恨言论。该数据集通过重新标注来自Twitter、Reddit和Wikipedia Talk Pages的内容创建，旨在区分挑衅和仇恨言论，提高自动检测有害内容的准确性。数据集的应用领域包括社交媒体内容过滤，旨在解决有害内容自动检测的问题。

The dataset 'Trawling for Trolling' was developed by the International Institute of Information Technology, India. It contains 12,490 samples, which are categorized into 5 categories: normal, profanity, provocation, defamation, and hate speech. This dataset is constructed by re-annotating content sourced from Twitter, Reddit, and Wikipedia Talk Pages. Its core goal is to distinguish between provocation and hate speech, and improve the accuracy of automated harmful content detection. The application fields of this dataset include social media content filtering, and it is designed to address the challenges in automated harmful content detection.