Arabic Offensive Language Classification on Twitter

Social media users often employ offensive language in their communication. Detecting offensive language on Twitter has many applications ranging from detecting/predicting conflict to measuring polarization. In this paper, we focus on building effective offensive tweet detection. We show that we can rapidly build a training set using a seed list of offensive words. Given the automatically created dataset, we trained a character n-gram based deep learning classifier that can effectively classify tweets with F1 score of 90%. We also show that we can expand our offensive word list by contrasting offensive and non-offensive tweets.