Background: Indonesian online discourse increasingly circulates emotion, irony, and public sentiment through civic complaint, protest symbolism, electoral distrust, misinformation correction, and transnational fandom conflict, making polarity-based sentiment analysis insufficient for culturally situated interpretation. Objective: This study examines how affective meaning is organized across registers and how irony, sarcasm, and pragmatic reversal reshape computational readings of Indonesian public sentiment. Method: This study applies corpus-based computational discourse analysis to 22 publicly verifiable corpus items, combining register-sensitive affect coding, irony auditing, and model–discourse misalignment analysis. Results: Civic-service complaint forms the largest affective register, showing that technical or bureaucratic failure is often transformed into public judgment of institutional competence. Irony and sarcasm operate as pragmatic filters that destabilize literal sentiment, especially through mock praise, rhetorical alarm, humorous ridicule, and group-directed antagonism. Model–discourse misalignment appears most strongly where sentiment depends on register, context, multimodal symbolism, or culturally embedded inference rather than lexical polarity alone. Implication: These findings suggest that Indonesian affective modeling requires interpretive validation alongside computational classification. Novelty: This study contributes a register-sensitive framework that integrates affect, irony, sentiment, and model-error interpretation within Indonesian computational discourse analysis.
Copyrights © 2026