Beyond Hostile Linguistic Cues: The Gravity of Online Milieu for Hate Speech Detection in Arabic


Religious Hate speech poses grave dangers for the cohesion of a democratic society, the protection of human rights and the rule of law. While previous work has shown that linguistic features can be effectively used for text categorization in Arabic, employing information coming from userssocial networks has not yet been explored for such complex user characteristics. Systems relying on language information tend to have low precision because they tend to rely on messages containing particular terms indicating hate speech. In this paper, we study the novel problem of exploiting social context for detection of religious hate speech in Arabic tweets, given information extracted from their online milieu by learning a low-dimensional vector representation of users.

In Proceedings of the 30th ACM Conference on Hypertext and Social Media