What does this imply? examining the impact of implicitness on the perception of hate speech

Abstract

We analyze whether implicitness affects human perception of hate speech. To do so, we use Tweets from an existing hate speech corpus and paraphrase them with rules to make the hate speech they contain more explicit. Comparing the judgment on the original and the paraphrased Tweets, our study indicates that implicitness is a factor in human and automatic hate speech detection. Hence, our study suggests that current automatic hate speech detection needs features that are more sensitive to implicitness.