Tracing Semantic Variation in Slang

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 1299 - 1313 December 7-11, 2022 ©2022 Association for Computational Linguistics Tracing Semantic Variation in Slang Zhewei Sun1, Yang Xu1,2 1Department of Computer Science, University of Toronto, Toronto, Canada 2Cognitive Science Program, University of Toronto, Toronto, Canada {zheweisun, yangxu}@cs.toronto.edu Abstract The meaning of a slang term can vary in dif- ferent communities. However, slang seman- tic variation is not well understood and under- explored in the natural language processing of slang. One existing view argues that slang se- mantic variation is driven by culture-dependent communicative needs. An alternative view fo- cuses on slang’s social functions suggesting that the desire to foster semantic distinction may have led to the historical emergence of community-specific slang senses. We explore these theories using computational models and test them against historical slang dictionary en- tries, with a focus on characterizing regularity in the geographical variation of slang usages attested in the US and the UK over the past two centuries. We show that our models are able to predict the regional identity of emerging slang word meanings from historical slang records. We offer empirical evidence that both commu- nicative need and semantic distinction play a role in the variation of slang meaning yet their relative importance fluctuates over the course of history. Our work offers an opportunity for incorporating historical cultural elements into the natural language processing of slang. 1 Introduction Slang is a type of informal language commonly used in both day-to-day conversations and online written text. The pervasiveness of slang has gen- erated increasing interest in the natural language processing (NLP) community, with systems pro- posed for automatic detection (Dhuliawala et al., 2016; Pei et al., 2019), generation (Sun et al., 2019, 2021), and interpretation (Ni and Wang, 2017; Sun et al., 2022) of slang. However, these existing ap- proaches do not account for the semantic variation of slang among different groups of users—a defin- ing characteristic of slang which distinguishes it from conventional language (Andersson and Trudg- ill, 1992; Mattiello, 2005; Eble, 2012). Figure 1 Shared 78% US 10% UK 11% Conventional Shared 55% US 23% UK 22% Slang Figure 1: Distribution of regional identities among sense entries found in the English Wiktionary. See Ap- pendix A for the detailed experimental setup. beast 1954: A fast car. 1982: Subway #2 of NYC. 1997: Excellent. 1837: An unpleasant person. 1877: A sexual offender. 1898: A bicycle. 2011: An outstanding example. “You’re a beast, man. You nailed that sucker.” Figure 2: Illustration of semantic variation in the slang word beast, with senses recorded in American and British English respectively. We develop slang semantic variation models to trace the regional identity of a new emerging slang sense given its historical meanings and usages from different regions. shows a tally of Wiktionary entries confirming that semantic variation is much more prevalent in slang compared to conventional language. Here we de- velop a principled computational approach to inves- tigate regularity in the semantic variation …

Content truncated to 500 words