<![CDATA[The 2024 Indonesian presidential election witnessed heightened political discourse on social media, accompanied by an alarming rise in misinformation. This study explores the use of lexicon augmented sentiment analysis as a method to detect hoax content in electoral conversations across Twitter, TikTok, and Meta platforms. By combining sentiment polarity analysis with weak supervision and partial manual validation, we developed a hybrid model tailored to Bahasa Indonesia. Using around 50,000 social media posts combined with a verified hoax index from MAFINDO, we examined how sentiment changed over time within political hashtags. We found that sentiment sharply declined after major events like debates and result announcements. Importantly, posts with very negative tone were 3–9 times more likely to contain misinformation, with 18% directly matching confirmed hoaxes. The hybrid model improved classification accuracy from 64% to 78%, showing its practical potential. The results confirm that sentiment polarity particularly extreme negativity can serve as a leading indicator for misinformation outbreaks. By aligning lexicon based sentiment scores with external verification sources, this framework enables scalable and semi automated detection of political hoaxes in low resource language settings. Ethical considerations in data handling, platform compliance, and demographic inclusivity are emphasized throughout the methodology. This research contributes to computational political analysis by validating a practical, replicable model for electoral misinformation detection. Future work should extend toward multimodal detection, real time dashboards, and participatory collaborations with fact checkers and regulatory bodies.]]>
Copyrights © 2024
