Analisis Eksperimental Kinerja Transformers, VADER, dan Naive Bayes dalam Analisis Sentimen Teks Bahasa Indonesia: Studi Kasus Komentar Terkait Judi Online

Abstract

Sentiment analysis is a subfield of Natural Language Processing (NLP) that focuses on detecting and classifying opinions expressed in textual data. In the digital social context, the increasing volume of public comments related to online gambling in Indonesia highlights the need to map public perception. This study aims to conduct an experimental analysis of the performance of three popular sentiment analysis approaches: VADER (Valence Aware Dictionary and sEntiment Reasoner), Naive Bayes, and Transformers-based models, specifically on Indonesian-language text. The dataset consists of public comments from social media and digital platforms containing keywords related to online gambling. The research process involves text preprocessing, data labeling, model training (for Naive Bayes and Transformers), and performance testing. Evaluation metrics include accuracy, precision, recall, and F1-score. The experimental results show that the Transformers model (using IndoBERT) achieves the highest performance in terms of accuracy and generalization ability, while VADER performs less optimally due to its limitations in understanding Indonesian linguistic context. Naive Bayes demonstrates moderate and consistent performance but lacks the capability to capture complex contextual meanings. These findings contribute to selecting appropriate sentiment analysis methods for non-English languages and support the development of more accurate public opinion detection systems in the future