The grammar of fake news: Evidence from English and Russian

Abstract: In this talk, we present preliminary findings from the Fakespeak project (funded by The Research Council of Norway) on the grammatical differences between fake and genuine news in English and Russian. Grammatical features are well suited as input for automatic fake news detection systems because they can be easily identified and because, in linguistics, they have been shown to reveal systematic differences between registers of language use. The methodological approach is Multidimensional Analysis, a robust corpus-based register analysis of grammatical features, which originated in English. The approach takes as its starting point large corpora of fake and genuine news, which in Fakespeak we have collected and classified based on the veracity labels of fact-checkers in English (PolitiFact) and Russian (Provereno.media). We start the talk by presenting the methodological approach and its applications in English register studies. Then, we present the preliminary findings in Russian, and, finally, we lay out our future plans for extending the studies both in English and Russian.