<![CDATA[YouTube, as a video-sharing platform, has become a public interaction space rich in opinions regarding online gambling issues in Indonesia. However, large-scale manual sentiment analysis is difficult due to the high data volume and local language nuances. This study aims to develop a sentiment classification model for Indonesian-language YouTube comments using Natural Language Processing (NLP) techniques to understand public perceptions of the online gambling phenomenon. Data of 3,000 comments were collected from YouTube videos related to online gambling through the YouTube Data API in Indonesia. All data were manually annotated by three annotators (kappa 0.85) into three sentiment classes (positive, negative, neutral) along with relevance, then divided into 80% training and 20% testing. Pre-processing included case folding, text cleaning, tokenization, stopword removal, stemming, lemmatization, and slang normalization. Models tested included Naive Bayes, Support Vector Machine (SVM), Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and IndoBERT. Evaluation using accuracy, precision, recall, and F1-score metrics showed IndoBERT achieved the best performance with 91.67% accuracy, 90% precision (negative class), 95% recall (negative class), and 91.66% F1-score. This research contributes to understanding public attitudes toward online gambling and the development of an adaptive sentiment classification system for the Indonesian language.]]>
