Homophobic hate speech in Italian tweets: Contextual cues of offensiveness

Social media channels have become omnipresent tools for communicating, sharing knowledge, and establishing communities, but also places where all sorts of hate speech comes to the surface. This chapter contributes to the body of work on online hate speech towards sexual orientation minorities and takes some initial steps towards a quantitative variationist sociolinguistic study of homophobic language in Italian social media. By exploring the different lexicalisations (i.e. near-synonyms) available in Italian for expressing or referring to the concept of "homosexual man" on X (formerly known as Twitter), this study provides both a descriptive and a methodological contribution. For this purpose, we created a dataset of 3000 manually annotated tweets in which at least one of the 33 lexicalisations of the concept "homosexual man" was recorded, and which we annotated for linguistic and stylistic factors that favour the perception of an expression as offensive. Our study shows that the interaction of lexical variation and explicit and implicit contextual cues, such as irony and dialect use, is significant in determining the degree of offensiveness of tweets. Furthermore, this study provides further evidence in favour of using social media as a laboratory for mapping language variation on a large scale, and for reflecting on the refinement of semi-automatic annotation of linguistic, stylistic, and social variables in written social media language.