Expressivity and Computational Linguistics

Abstract This chapter provides an overview of computational approaches to expressivity. It argues that computational linguistics has addressed expressivity in a piecemeal way by analyzing individual phenomena that include expressive items. However, the question of the differences between expressive and descriptive meaning has not been systematically studied in computational linguistics. One possible exception is the area of hate speech and its detection, a subfield where computational linguistic research has expended significant effort on analyzing expressive items. This chapter argues that expressivity can often be seen as a lexical semantic feature, and introduces the main approaches to distributional semantics. It is further shown that current distributional models insufficiently represent the expressive/descriptive distinction, using intensifiers as a case study. Finally, the chapter describes paths via which computational linguistics may contribute more substantially to research on expressivity in the future.