Hate speech, both written and spoken, is a growing source of concern as it often discriminates societal minorities for their national origin, sexual orientation, gender or disabilities. Despite its destructive power, hardly anything is known about whether there are cross-linguistic mechanisms and acoustic-phonetic characteristics of hate speech. For this reason, our experiment analyzes the implicit prosodies that are caused by written Twitter and Facebook hate-speech items and made phonetically "tangible" through a special, introspective reading-aloud task. We compare the elicited (implicit) prosodies of Danish and German speakers with respect to f0, intensity, HNR, and the Hammarberg index. While we found no evidence for a consistent hate-speech-specific prosody either within or between the two languages, our results show clear prosodic differences associated with types of hate speech and their targeted minority groups. Moreover, language-specific differences suggest that – compared to Danish – German hate speech sounds more expressive and hateful. Results are discussed regarding their implications for the perceived severity and the automatic flagging and deletion of hate-speech posts in social media.
|Titel||Proc. Interspeech 2020|
|Forlag||International Speech Communication Association (ISCA)|
|Publikationsdato||27. okt. 2020|
|Status||Udgivet - 27. okt. 2020|
|Begivenhed||Interspeech 2020 - Shanghai, Kina|
Varighed: 25. okt. 2020 → 29. okt. 2020
|Periode||25/10/2020 → 29/10/2020|