RTAnews: A Benchmark for Multi-label Arabic Text Categorization

RTAnews dataset is a collections of multi-label Arabic texts, collected form Russia Today in Arabic news portal. It consists of 23,837 texts (news articles) distributed over 40 categories, and divided into 15,001 texts for the training and 8,836 texts for the test. The original dataset (without preprocessing), a preprocessed version of the dataset, versions of the dataset in MEKA and Mulan formats, single-label version, and WEAK version all are available. For any enquiry or support regarding the dataset, please feel free to contact us via bassalemi at gmail dot com