Comparative Analysis of Online Political Discourse Across Languages and Platforms

Thesis Abstract

This study investigates how political discourse online varies across languages and digital platforms, addressing the problem of cross-linguistic and cross-platform polarization, mis/translation challenges, and modality differences in political persuasion that hinder comparative understanding for researchers and policymakers. The aim is to identify linguistic markers, discourse strategies, and platform-induced constraints that shape political communication in multilingual, multi-platform environments, with specific objectives to (1) compare lexical choices, sentiment orientation, and messaging strategies across English, Spanish, and Mandarin posts; (2) examine syntactic and discourse-pragmatic patterns linked to platform features (microblogging, forum threads, and video comment sections); (3) assess cross-language translation effects on discourse cohesion and stance; (4) evaluate how audience engagement metrics correlate with discourse styles; and (5) develop a cross-language, cross-platform analytic framework for monitoring online political rhetoric. The study employs a mixed-methods design, combining quantitative corpus analysis with qualitative thematic analysis, underpinned by the Social Identity Theory and the Rhetorical Structure Theory as the theoretical lenses. The population comprises publicly available political posts and comment threads from major platforms—Twitter/X, Reddit, and YouTube comment sections—during a six-month political event window in 2023–2024. A stratified sample includes 1,200 posts (400 per language English, Spanish, Mandarin) and corresponding 6,000 comments (2,000 per language) collected via platform APIs and archived datasets, balanced for topic diversity (campaign announcements, policy debates, and opinionated responses). Data collection instruments include a multilingual parallel corpus tool for consistent alignment of translations and a coding schema derived from the Discourse Transitivity framework and polarity annotation guidelines, complemented by a platform feature taxonomy capturing character limits, threading, and annotation affordances. Validity and reliability are ensured through intercoder agreement (Cohen’s kappa target ? .80 for qualitative coding) and triangulation across linguistic experts, platform engineers, and political communication researchers. The primary analysis involves multi-layered techniques (i) corpus-based quantitative analysis using word embeddings (BERT multilingual) and LIWC-style lexicons to quantify sentiment, stance, and modality; (ii) cross-language regression analyses to test the influence of language and platform on discourse features such as hedging, assertiveness, and evaluative language; (iii) ANOVA to examine differences across platforms within each language and post-hoc contrasts for pairwise comparisons; (iv) thematic analysis of a purposive subsample (n = 180 posts and 900 comments) to extract coherent discursive strategies and rhetorical patterns; and (v) network analysis of user interactions to map influence patterns by language and platform. The anticipated findings include distinct cross-language divergences in stance-taking and evaluative framing, with Mandarin posts showing higher use of collectivist framing and indirect stance markers; English and Spanish discourse exhibiting more explicit polarity and direct-address strategies; platform-specific constraints shaping discourse length, immediacy, and narrative coherence, with Twitter/X favoring concise argumentative moves, Reddit enabling extended argument chains, and YouTube comments privileging evaluative discourse and community surveillance cues. The study expects translation-related attenuation of stance clarity in multilingual threads, moderated by platform affordances and community norms. Theoretical contributions include integration of Social Identity Theory with Rhetorical Structure Theory to explain how multilingual communities construct in-group/out-group distinctions through discourse moves across platforms; methodologically, the research offers a replicable cross-language, cross-platform analytic framework combining corpus linguistics, psycholinguistic probing, and network analytics. The practical implications span policy communication, platform moderation, and digital literacy, enabling stakeholders to monitor cross-linguistic political narratives, anticipate mis/disinformation risks, and tailor multilingual communication strategies. The study concludes that language and platform jointly shape online political discourse, with significant implications for intercultural understanding, democratic participation, and global political communication governance. Recommendations include developing standardized multilingual monitoring dashboards, investing in cross-linguistic translation quality controls for political content, and promoting platform-specific best practices to foster transparent, accountable political dialogue.