Online Library

This paper presents a survey on hate speech detection. Given the steadily growing body of social media content, the amount of online hate speech is also increasing. Due to the massive scale of the web, methods that automatically detect hate speech are required. Our survey describes key areas that have been explored to automatically recognize these types of utterances using natural language processing. We also discuss limits of those approaches.

VOX-Pol Blog post.

The far right is increasingly relying on visual and less extreme online communication, for instance by using memes, to strategically mainstream their ideology. The use of humor in particular renders their communication more relatable to a mainstream audience. However, little is known about the actual impacts of the different content characteristics they employ to become more appealing, in particular on less moderated platforms that function as safe online spaces for extremist ideology and contents. To fill this gap, we conducted a manual quantitative content analysis of 1,200 memes distributed within German-language far-right Telegram channels in 2020 and 2021, concentrating on humor and several content-related factors to analyze their impact on meme reach. The results demonstrate that memes with extreme far-right narratives and memes with humor received fewer views than others, but that memes with both far-right narratives and humor had a significantly increased reach. The findings highlight the mainstreaming potential of humor, particularly when used to mask extreme content that would otherwise be less appealing.

Although many social media users have reported encountering hate speech, differences in the perception between different users remain unclear. Using a qualitative multi-method approach, we investigated how personal characteristics, the presentation form, and content-related characteristics influence social media users’ perceptions of hate speech, which we differentiated as first-level (i.e. recognizing hate speech) and second-level perceptions (i.e. attitude toward it). To that end, we first observed 23 German-speaking social media users as they scrolled through a fictitious social media feed featuring hate speech. Next, we conducted remote self-confrontation interviews to discuss the content and semi-structured interviews involving interactive tasks. Although it became apparent that perceptions are highly individual, some overarching tendencies emerged. The results suggest that the perception of and indignation toward hate speech decreases as social media use increases. Moreover, direct and prosecutable hate speech is perceived as being particularly negative, especially in visual presentation form.