Sentiment Analysis of Instagram Comments for Monitoring Personal Branding of YBM Brilian Scholarship Recipients, Regional Office, Makassar

This study focuses on the implementation of the Multilingual BERT (mBERT) architecture combined with a Long Short-Term Memory (LSTM) model to classify Instagram comments into positive, negative, and neutral sentiments. The primary objective is to support the monitoring of personal branding among recipients of the Bright Scholarship managed by the Baitul Mall BRILiaN Foundation (YBMRILiaN) at the Makassar Regional Office. The experimental results indicate that mBERT is capable of effectively analyzing sentiment from Instagram comments on scholarship awardees from Hasanuddin University and UIN Alauddin Makassar. Using a sample of 10 awardees, the model demonstrates a consistent increase in accuracy across epochs, achieving an average accuracy of 63.87% and a peak accuracy of 73.18% for Awardee 10, with a corresponding loss value of 1.094. These findings highlight the potential of this approach to assist scholarship organizers in systematically evaluating the personal branding of awardees on social media. Moreover, the analysis identifies one awardee whose personal branding performance may require further consideration regarding scholarship eligibility.