Bandung, Indonesia — A recent study published in Jurnal Otomasi Kontrol dan Instrumentasi (JOKI) presents an innovative approach to understanding public reactions toward the 2025 UTBK (Ujian Tulis Berbasis Komputer) announcement through artificial intelligence and multi-label emotion analysis.
Conducted by researchers from Institut Teknologi Sumatera (ITERA), the study explores how Indonesian social media users expressed their emotions on X (formerly Twitter) following one of the country's most anticipated educational events. Rather than limiting analysis to positive or negative sentiment, the researchers employed a multi-label emotion classification framework capable of detecting multiple emotional states simultaneously within a single post.
The research introduces a comprehensive calibration pipeline built upon the IndoBERTweet language model. The framework combines posterior probability calibration, precision-oriented threshold optimization, prevalence rate targeting, and lexicon-aware score enhancement to improve the detection of minority emotional categories and reduce prediction bias caused by data imbalance.
Using more than 3,500 Indonesian-language tweets related to the UTBK announcement, the system successfully mapped diverse emotional responses, including anticipation, joy, trust, sadness, fear, and disappointment. The findings reveal the complex emotional landscape experienced by students and the broader public during the announcement period.
According to the authors, social media emotion mapping can provide valuable insights for educational institutions, policymakers, and communication strategists seeking to better understand public responses to large-scale national events. The proposed methodology is also designed to be model-agnostic, enabling future applications across various domains such as public policy analysis, disaster communication, healthcare monitoring, and social media intelligence.
This study demonstrates how modern natural language processing, machine learning calibration techniques, and Indonesian language resources can be combined to generate more reliable and interpretable emotion analytics from real-world social media data.
The article, titled "Mapping Public Emotions Regarding the 2025 UTBK Announcement in Indonesia: A Multi-Label Approach with Targeted Calibration," appears in JOKI Volume 18 Issue 1 (2026) and contributes to the growing field of intelligent text analytics, emotion-aware AI systems, and social media monitoring technologies.
Keywords: Artificial Intelligence, Emotion Analysis, Multi-Label Classification, IndoBERTweet, Social Media Analytics, Educational Data Mining, Natural Language Processing, Indonesia.