Analisis Sentiment Netizen Indonesia pada Berita Viral di Akun Instagram Folkative Menggunakan Model Indonesian Bidirectional Representation Encoder from Transformers (IndoBERT)

This study aims to analyze the sentiments of Indonesian netizens toward viral news posted on the Instagram account Folkative during the period from January to July 2025, with one viral news item selected from each month. The analysis was conducted to identify patterns of public opinion, which were classified into four categories: positive, negative, almost positive, and almost negative, using the IndoBERT model. The research data were collected through web scraping using Instans Data Scraper based on criteria requiring posts to have more than 200,000 likes and 1,000 comments.The pre-processing stage consisted of case folding, removal of non-text characters, filtering, and duplicate elimination. IndoBERT was utilized to perform sentiment classification and evaluate its performance through accuracy, precision, recall, and F1-score metrics. The use of IndoBERT for classifying the four sentiment categories revealed that the model was capable of recognizing linguistic patterns containing emotions, sarcasm, and complex opinions.The findings also provide an overview of the dynamics of Indonesian public opinion toward viral content, demonstrating IndoBERT’s strong potential as an accurate sentiment analysis solution. With its bidirectional context understanding, the model is highly suitable for analyzing Instagram comments, which are generally informal. Therefore, this study adopts the IndoBERT model to analyze public sentiment toward viral news posted on the Instagram account Folkative during the January–July 2025 period.