Analyzing Public Sentiment on Indonesia's Constitutional Court Post-2024 Election Ruling: Insights from Appraisal Theory and Data Mining

Abstract

This study examines public sentiment toward Indonesia’s Constitutional Court (Mahkamah Konstitusi, MK) following its 2024 regional election ruling. Using sentiment analysis and Martin and White’s (2005) Appraisal Theory, the research investigates emotional, evaluative, and dialogic patterns in public discourse through YouTube comments. A mixed-method approach was adopted by combining qualitative appraisal interpretation with quantitative categorization and machine learning-based sentiment classification, all conducted using the Orange data mining application. Orange was chosen for its visual programming interface, ease of integration between linguistic theory and machine learning workflows, and accessibility for researchers working across disciplines. From 4,010 YouTube comments, 223 relevant entries were filtered and analysed according to the three domains of Appraisal Theory: Attitude (affect, judgment, appreciation), Engagement (monogloss and heterogloss), and Graduation (force and focus), enabling a structured evaluation of public responses. Three machine learning models were employed for sentiment classification within Orange: Naive Bayes, for its speed and efficiency in text classification; Logistic Regression, for its interpretability and robust baseline performance; and Neural Network, for its ability to capture nuanced emotional expressions. Among these, the Neural Network achieved the highest performance (AUC: 0.958; F1 score: 0.853), followed by Logistic Regression (AUC: 0.931; F1: 0.807), and Naive Bayes (AUC: 0.925; F1: 0.802). Each model offered distinct strengths: Neural Network revealed deeper emotional intensity, Logistic Regression emphasized positive affect, and Naive Bayes captured dominant monoglossic tendencies in discourse. The findings reveal a predominance of neutral and moderately positive sentiments, with joy, fear, surprise, and dissatisfaction emerging as key affective responses. The integration of Appraisal Theory and sentiment modeling through Orange demonstrates a systematic and scalable method for interpreting public discourse in digital environments. This research contributes methodologically by bridging qualitative linguistic analysis with accessible data mining tools, and substantively by offering insight into how digital publics engage with constitutional authority. It advances the literature on institutional trust by illustrating how social media serves as a platform for democratic evaluations of judicial decisions.