<![CDATA[The Sekolah Rakyat program is a strategic Ministry of Social Affairs initiative requiring continuous evaluation through public perception monitoring. This study employs Aspect-Based Sentiment Analysis (ABSA) using the IndoBERT transformer model via a two-stage approach: aspect-opinion extraction using BIO labeling (Token Classification) and sentiment polarity determination (Sequence Classification). A dataset of 14,787 entries from Platform X underwent systematic preprocessing and was trained using an 80:20 stratified split to ensure label balance. Model performance demonstrated high reliability, achieving 86% accuracy and stable F1-scores. Collectively, the analysis identified 8,454 neutral, 4,062 positive, and 2,271 negative sentiments. The results reveal that educational aspects, specifically regarding students, are the primary focus of public discourse, dominated by neutral sentiment. These findings confirm that Deep Learning-based approaches provide granular insights into policy effectiveness, serving as an accurate decision-support instrument for the government to evaluate educational policies comprehensively based on data-driven evidence.]]>
Copyrights © 2026
