OPINION MINING AND SENTIMENT ANALYSIS OF THE TEACHER PROFESSIONAL EDUCATION APPLICATION ECOSYSTEM BASED ON THE ISO/IEC 25010 STANDARD

Abstract

The digital transformation in the government services sector (GovTech), particularly within Indonesia's public education ecosystem, has led to the mandatory adoption of digital applications for the Teacher Professional Education (PPG) program. While designed to improve bureaucratic efficiency and pedagogical competence, grassroots adoption frequently encounters severe technical barriers and server instability, triggering "technostress" among educators. Traditional software evaluation methods, which rely on quantitative metrics or macro-level sentiment analysis, fail to capture the technical granularity of user complaints. This study aims to design a hybrid analytical architecture that automatically mines, processes, classifies, and dissects the sentiments of hundreds of thousands of user reviews, contextualising them within the ISO/IEC 25010 software quality standard. Utilising the Knowledge Discovery in Databases (KDD) framework, this research employed automated web scraping, multi-layered text preprocessing, and a comparative classification modelling approach. The state-of-the-art IndoBERT model was compared against a traditional Support Vector Machine (SVM) baseline, supplemented by Latent Dirichlet Allocation (LDA) for unsupervised topic modelling. The results demonstrate IndoBERT's computational superiority, achieving 95.2% accuracy and 95.4% F1 score, significantly outperforming the SVM. An aspect-based diagnostic mapping revealed that critical system failures were predominantly rooted in performance efficiency (32.5%) and compatibility (28.2%) deficiencies. The study concludes that the success of e-government applications cannot be measured solely by feature quantity but requires stringent algorithmic stability, cross-platform interoperability (including iOS), and optimised resource utilisation. We recommend transitioning from conventional surveys to an AI-driven, proactive sentiment telemetry dashboard to mitigate software anomalies in real time.