Banglalish on social media: A corpus-based study of Bangla–English code-mixing across four platforms in Bangladesh

Abstract

This article reports a 12-month, 120,000-token corpus study of Bangla–English codemixing in Bangladesh’s social-media discourse across Facebook, Twitter, YouTube, and Instagram. Anchored in variational sociolinguistics and contact linguistics, we operationalise code-mixing using the Matrix Language Frame (MLF) model and related structural diagnostics (insertion vs. alternation; borrowing vs. switching), and we quantify platform effects on mixing frequency and structure. Findings indicate stable insertional patterns with Bangla as the matrix language; productive hybrid verb constructions (e.g., comment korchi, share korbo); strong preferences for Bangla determiners with English content nouns (e.g., ei app, amar phone); and domain-specific borrowing of digitalculture vocabulary. Platform affordances condition mixing: Facebook supports longer mixed turns, while Twitter compresses mixing and favours hashtag innovations; YouTube concentrates emphasis and evaluation in comment bursts; Instagram integrates visualtextual cues with searchable, multilingual tags. Demographic correlates (age, education, gender) are treated as tendencies rather than fixed categories. The article contributes (i) a transparent, replicable coding scheme for code-mixing in South Asia, (ii) cross-platform evidence of code-mixing as structured, not ad hoc, and (iii) research/pedagogical/policy implications of treating code-mixing as a regular, rule-bound resource in Bangladesh’s digital communication ecology.