The dataset consists of comments that are in some of the top news stories in 2019, obtained from several online news/forum, such as: kompas, kaskus, and detik. The labeling process is carried out by a total of 10 annotators and each comment is annotated by 3 annotators. Each comment was labeled with one of the following labels: not abusive, abusive but not offensive, abusive and offensive.
wenliangdai
commented
2 years ago
NusaCatalogue: https://indonlp.github.io/nusa-catalogue/card.html?id_abusive_news_comment