Sentiment and Discourse Analysis of Philippine Political Conversations on Social Media using Natural Language Processing
Year: 2026
Type: conference-paper
Abstract: Social media has become a major platform for political discourse in the Philippines, where online discussions often reflect public opinion, political polarization, and civic engagement. However, the large volume and complexity of social media conversations make traditional discourse analysis difficult, creating a need for automated and scalable analytical approaches. To address this issue, the study applied Natural Language Processing (NLP) techniques, specifically sentiment analysis and topic modeling, to analyze Facebook comments related to Philippine political discussions. Transformer-based models and topic modeling approaches were utilized to identify dominant sentiments and thematic structures present in the dataset. The findings revealed that most Facebook comments expressed positive sentiment, although negative and neutral sentiments were significant, indicating diverse public perspectives. Topic modeling identified Governance, Leadership, and Criticism as dominant themes associated with political institutions, leadership performance, and public accountability. Furthermore, the integrated sentiment-topic analysis showed that discussions related to Governance and Criticism generated more negative reactions, while Leadership and Economy-related discussions elicited engagement that is more positive. These findings demonstrate the relationship between thematic salience and public sentiment in online political discourse. The study highlights the importance of NLP techniques in understanding large-scale political discussions on social media. The findings may assist government institutions, policymakers, and researchers in examining public opinion, political participation, and online discourse behavior. Future studies may improve analysis accuracy through enhanced sarcasm detection, multilingual preprocessing, and broader social media datasets.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203495680