Our project approaches the definition of hate-speech from the addressee's perspective (i.e., the perception of written and spoken language) and proceeds from objectified empirical evidence to common linguistic denominators of the phenomenon. Our aim is helping companies and the society to better detect hate speech and distinguish its various subtypes.
Nowadays, the term “hate speech” is frequently used for any kind of harsh comments [1, 2]. Definitions either refer to subjective intuitions or general stylistic criteria in combination with the sender's intention [3]. However, wouldn't it make sense to approach a hate-speech definition from the addressee's perspective and proceed, on this basis, from objectified empirical evidence to common linguistic denominators of the phenomenon? Taking this approach, we use original German hate speech data as stimuli in a perception experiment, enriched by linguistic expressions (e.g., irony, holocaust-relations) that often co-occur with hate speech comments. Our aim is helping companies and the society to better detect hate speech and, possibly, distinguish its various subtypes. So far, 18 German native speakers read, listened to (or vice versa) and evaluated the target sentences (written-spoken: 11 participants; spoken-written: 7 participants). Judgments were made by clicking into a white rectangle, describing hate speech in terms of combined attributes: acceptability on the x-axis and need for consequences on the y-axis. Results for both written and spoken language identify holocaust and ironic stimuli as the rating’s extreme points, whereas all other types are judged similarly. While holocaust stimuli are rated as absolutely inacceptable with a strong need for consequences, irony is the most acceptable type where consequences are less important due to its ambiguity. If participants are first presented with spoken language, the difference between holocaust stimuli and all other types is most apparent. Further participants will give a clearer picture of the stimuli distribution. This empirical study aims at filling the gap between the linguistic and communicative mechanisms underlying the expression of hate speech and how ordinary language users perceive particular hate speech structures.Understanding how prosody shapes hate speech perception in spoken language will help to better define and analyse it, also in the written domain.
