Hate speech annotation: Analysis of an Italian twitter corpus