Directed Digital Hate

Politicians, journalists and public figures are often exposed to hate speech in various digital environments, especially in different discussion forums and in comment fields. Detecting hate automatically is a difficult task since hate can be expressed in many different ways. In this paper we have developed a method to measure hate directed at politicians using a combination of natural language processing and automated reasoning. Our method is adapted to work on Swedish, although in general it is language independent. We have tested our method in a study where we analyze hate directed at six Swedish politicians. The results shows that our method has a fairly high precision but a low recall compared to a manual assessment.