Proceedings of the Annual Meeting of the Cognitive Science Society

Slang is a common device for expressivity in natural lan-guage. While slang has been studied extensively as a socialphenomenon, its cognitive bases are not well understood. Weformulate the processes of slang generation as a categoriza-tion problem. We explore a set of cognitive models of catego-rization that recommend slang words based on intended refer-ents of the speaker beyond the existing senses of words. Wetest these models against a large repertoire of slang sense def-initions from the Online Slang Dictionary and show that thecategorization models predict slang word choices substantiallybetter than chance, without explicit consideration of externalsocial factors. We also show that words similar in existingsenses tend to extend to similar novel slang senses, reflecting aprocess of parallel semantic change. Our work helps to groundtheories of slang in cognitive models of categorization and pro-vides the potential for machine processing of informal naturallanguage.