ComMA Dataset V0.2

ComMA Dataset V0.2是由潘林格语言处理有限责任公司和Dr Bhimrao Ambedkar大学合作创建的多语言数据集，专注于标记社交媒体话语中的攻击性和偏见。数据集包含15,000条注释评论，涵盖Meitei, Bangla, Hindi, 和 Indian English四种语言，主要从YouTube, Facebook, Twitter 和 Telegram收集。数据集的创建过程涉及从具有社会政治、宗教或文化争议的视频和帖子中收集评论，并通过细致的注释过程来标记不同类型的攻击性和偏见，包括性别偏见、宗教不容忍、阶级/种姓偏见和种族/民族偏见。该数据集旨在帮助开发自动识别社交媒体中攻击性和偏见的系统，以减少其对社会交流和整体社会的负面影响。

ComMA Dataset V0.2 is a multilingual dataset co-created by Panlingua Language Processing Co., Ltd. and Dr. Bhimrao Ambedkar University, focusing on annotating aggression and bias in social media discourse. The dataset contains 15,000 annotated comments across four languages: Meitei, Bangla, Hindi, and Indian English, and was primarily collected from YouTube, Facebook, Twitter and Telegram. The dataset creation process involves collecting comments from videos and posts related to socio-political, religious, or cultural controversies, followed by a meticulous annotation procedure to label various types of aggression and bias, including gender bias, religious intolerance, class/caste bias, and racial/ethnic bias. This dataset aims to facilitate the development of automated systems for identifying aggression and bias in social media, so as to mitigate their negative impacts on social communication and broader society.