Automated Detection of Gendered Linguistic Variation in Arab EFL Social Media Discourse Using Machine Learning and NLP

Abstract

This mixed method study examines gendered differences in the language used by Arabic speakers during online communication, a little researched phenomenon in the Arabic speakers’ paradigm compared to the western context. Applying the sociolinguistic model, the study tests the hypothesis that women use "standard/prestige” language forms as compared to the "vernacular/innovation” form used by men. This study analyzes three registers (standard, vernacular, innovative) in code-switched English and how these are created by Arab EFL users during communication on social networks like X/Twitter, Instagram, Snapchat, and TikTok in the context of the performative hybrid register of social media. A corpus equally representative of men and women users comprised the dataset for this study followed by coding for phonological, lexical, morphological, code-switching variants, and paralinguistic features such as emojis. Quantitative data was analyzed using Chi-square test, correlation analysis, and regression analysis to compare the distribution of variants by gender, age, education, and platform. Discourse analysis was applied to qualitative data to identify and examine the interpretive role of linguistic variants, such as humor, emphasis, and cosmopolitan self-presentation. The study results were illustrative with clear evidence of systematic gender differences in online language use, strongly intertwined with social and technological factors, implying that gender needs to be modeled as part of a larger "gestalt" of social and technological factors. The study adds to the field of sociolinguistic theory in the context of spoken Arabic dialects and their extension into digital EFL discourse.