Offensive Language Detection of Arabic Tweets Using Deep Learning Algorithm

Offensive language has become a common occurrence in Arabic social media. Toxic textual content can be prohibited more easily with the use of automatic offensive language identification technologies. In this study, a deep learning approach was used to detect offensive Arabic language. Three models were used: RNN with LSTM, RNN with BLSTM, and the SVM learning algorithm. A dataset of 7000 tweets with two attributes from the Egyptian dialect was used. After data preprocessing, a 6-fold cross-validation was used to train and test the data. The evaluation of the three models showed the suitability of the RNN models (with an accuracy of 95.6%) over the SVM.