<![CDATA[This study develops an Aspect-Based Sentiment Analysis (ABSA) system to assess public opinion on the Kurikulum Merdeka policy using data from platform X. The approach integrates web scraping, Indonesian linguistic-aware preprocessing, fine-tuning of IndoBERT, Naive Bayes, and an ensemble weighted voting strategy to extract aspect-level sentiment. The dataset consists of 345 tweets collected between October 20 and November 10. The findings indicate that the most frequently discussed aspects include general issues, teacher readiness, and student impact, with negative sentiment emerging in several categories. The model achieved an accuracy of 91.59%, with a macro F1-score of 86.43% and a weighted F1-score of 91.69%. The study also identifies data limitations, particularly tweets that are short or multimedia-based, which often result in neutral classifications. Future improvements may include expanding the dataset, enhancing annotation quality, and exploring newer transformer-based approaches.]]>
Copyrights © 2026
