Irony markers across Arabic varieties in sarcastic twitter discourse: a corpus-based study

Abstract

This study presents, to the best of my knowledge, the first large-scale, cross-varietal, corpus-based analysis of irony markers in Arabic sarcastic Twitter discourse. It examines 2,954 sarcastic tweets from the ArSarcasm-v2 dataset, using a mixed-method approach that combines computational detection via Python scripts with manual verification. In doing so, it identifies and analyzes five categories of irony markers—morpho-syntactic, typographic, direct, schematic, and trope-based—across Modern Standard Arabic (MSA), Egyptian Arabic, Gulf Arabic, and Levantine Arabic, using an established typology adapted to Arabic digital discourse. Statistical analyses ( χ ² and Cramer’s V) point to clear cross-varietal variation. In fact, morpho-syntactic markers dominated (83.6 %), indicating that irony in Arabic is often structurally encoded through exclamations, negation, and interjections. As might be expected in diglossic contexts, MSA favored direct cues such as hashtags, whereas Gulf and Levantine users relied more on typographic devices (for example, emojis and punctuation), and Egyptian sarcasm appeared distinctly dialogic and performative. Thematic coding showed that political commentary constituted about half of the dataset, followed by social and pop-cultural issues, each making up roughly a fifth. By adapting this typology to the linguistic and typographic constraints of Arabic, the study extends existing models of irony to a diglossic, digitally mediated context. The findings show that Arabic sarcasm on Twitter is not a fixed style but a dynamic performance shaped by linguistic variety, platform conventions, and sociocultural norms, underscoring the need for dialect-sensitive, context-aware models in both linguistic theory and computational sarcasm detection.