Trend AnalysisLinguistics & NLPMixed Methods

Discourse Analysis in Social Media Conversations: Language, Power, and Digital Public Spheres

Social media has become the dominant arena for public discourse, but analyzing its linguistic patterns requires new methods that bridge critical discourse analysis with computational NLP approaches.

By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.

Social media platforms have fundamentally altered the structure of public discourse. Conversations that once occurred in constrained, ephemeral settings, cafes, town halls, letters to editors, now unfold in persistent, searchable, algorithmically amplified digital spaces. For discourse analysis, a field traditionally focused on close reading of relatively small text samples, social media presents both an unprecedented opportunity and a methodological challenge. The opportunity lies in the sheer scale and accessibility of naturally occurring discourse data. The challenge lies in analyzing millions of texts while preserving the contextual sensitivity that makes discourse analysis theoretically valuable.

Why It Matters

The language of social media is not merely a reflection of social reality but an active force in constructing it. How issues are framed in online debates shapes public opinion, policy agendas, and individual beliefs. The linguistic mechanisms of this influence, including metaphor, presupposition, intertextuality, positioning, and stance-taking, are precisely what discourse analysis is designed to reveal. But traditional Critical Discourse Analysis (CDA), developed for analyzing newspaper editorials and political speeches, must adapt to handle the distinctive features of digital discourse: brevity, informality, multimodality, threading and reply structures, algorithmic curation, and the collapse of public and private registers.

The practical stakes are high. Understanding how discourse operates in digital spaces is essential for addressing misinformation, political polarization, mental health impacts of online communication, and the design of healthier digital public spheres.

The Science

Critical Discourse Analysis of Online Debates

Afrasiab et al. (2025) apply Fairclough's Critical Discourse Analysis framework to online debates, examining how power dynamics and ideological representations are constructed through language in digital spaces. Their analysis reveals that social media debates deploy distinctive discursive strategies: appeals to personal experience as authority, strategic use of hedging and intensification, in-group/out-group construction through pronoun choice, and the weaponization of reported speech (quoting opponents to reframe their positions). The study demonstrates that while the medium is new, many of the discursive mechanisms are continuations of longstanding rhetorical strategies adapted to the affordances and constraints of digital platforms. However, platform-specific features like character limits, threading structures, and emoji reactions create genuinely novel discourse patterns that require extending traditional CDA categories.

Computational Tools for Domain-Specific Discourse

Agarwala et al. (2025) present Muse-it, a specialized tool for analyzing music discourse on Reddit, demonstrating how NLP techniques can be adapted for domain-specific discourse analysis. The tool combines sentiment analysis, topic modeling, and network analysis to map the structure of conversational engagement around music. While the domain is specific, the methodological approach is generalizable: combining NLP's ability to process large volumes of text with discourse analysis's attention to context, community norms, and discursive practices. The tool reveals discourse patterns that would be invisible in either purely manual or purely automated analysis, such as how genre communities construct aesthetic values through specific evaluative vocabulary and how expertise is performed and contested in online music discussions.

Mental Health Discourse and Clinical Detection

Kasanneni et al. (2025) apply machine learning and deep learning to social media conversations for mental health diagnostics, treating discourse patterns as clinical signals. Their analysis of how individuals express psychological distress through language connects discourse analysis to clinical applications. The linguistic markers they identify, including shifts in pronoun use, changes in temporal reference, decreased linguistic complexity, and specific affective vocabulary patterns, have both computational value (as features for classification models) and theoretical significance (as evidence about how mental health conditions manifest in language production). The work raises important ethical questions about the boundary between discourse analysis for understanding and discourse surveillance for diagnosis.

NLP-Driven Discourse Tracking at Scale

Maarif et al. (2024) demonstrate how NLP can reveal discourse patterns in Twitter conversations about circular economy, using topic modeling, sentiment analysis, and network analysis to map how sustainability discourse evolves over time. Their approach reveals that public discourse around circular economy is fragmented into distinct thematic clusters that rarely interact, with emotional valence varying significantly across topics. Positive sentiment dominates discussions of innovation and business models, while negative sentiment concentrates around waste management and policy implementation. The study exemplifies the trend toward using computational methods to identify macro-level discourse patterns that emerge from millions of individual utterances.

Discourse Analysis Methods: Traditional vs. Computational

<
DimensionTraditional CDAComputational NLPIntegrated Approach
Scale10-100 textsMillions of textsNLP filtering + CDA deep analysis
Context sensitivityHighLowContext-aware NLP + expert interpretation
ReproducibilityLow (analyst-dependent)HighSystematic sampling + interpretive validation
Power/ideology analysisCore strengthEmerging (stance detection)NLP identifies patterns, CDA interprets them
MultimodalityEstablished methodsImage+text models emergingMultimodal NLP + discourse theory

What To Watch

The integration of large language models into discourse analysis is both promising and problematic. LLMs can annotate discourse features (stance, framing, speech acts) at scale with reasonable accuracy, potentially bridging the gap between close reading and big data. But LLMs also carry their own discursive biases, having been trained on text that embeds particular ideological positions. The recursive challenge of using language models shaped by discourse to analyze discourse is both a methodological puzzle and a theoretical opportunity. Additionally, the shift toward multimodal social media content (short-form video, image-text combinations, voice messages) demands discourse analysis methods that can handle non-textual semiotic resources, pushing the field toward computational multimodal discourse analysis.

Discover related work using ORAA ResearchBrain.

References (4)

[1] Afrasiab, M., Imran, S., & Sabih, S. (2025). The Language of Social Media: A Critical Discourse Analysis of Online Debates. SRA, 3(2).
[2] Agarwala, J., Paul, G., & Vardhan, N.H. (2025). Muse-it: A Tool for Analyzing Music Discourse on Reddit.
[3] Kasanneni, Y., Duggal, A., & Ragupathi, S. (2025). Effective Analysis of Machine and Deep Learning Methods for Diagnosing Mental Health Using Social Media Conversations. IEEE Trans. Computational Social Systems.
[4] Maarif, M.R., Pratama, A., & Hutahaean, M. (2024). Tweeting Circular Economy: Unveiling Current Discourse Through Natural Language Processing. Sustainable Development.

Explore this topic deeper

Search 290M+ papers, detect research gaps, and find what hasn't been studied yet.

Click to remove unwanted keywords

Search 7 keywords โ†’