Sentiment Analysis of Indonesian Temple Reviews Using Lexicon-Based Features and Stochastic Gradient Descent

TripAdvisor is an online platform where people can review tourist destinations. With so many reviews available, assessing public sentiment can be challenging. We conducted sentiment analysis on the review data from Indonesia's five most popular temples: Borobudur, Prambanan, Plaosan, Ijo, and Mendhut. Ratings on Tripadvisor reviews do not determine positive and negative sentences. Therefore, an automated method is necessary. Our research proposed using Lexicon-based and Stochastic Gradient Descent techniques for sentiment analysis. This study uses a lexicon-based because it can determine the weighting for the orientation of positive and negative sentiment words. SGD is used to build classification models because it can solve problems with large and complex datasets. The evaluation results showed that our method can significantly increase the model's performance. The result without a lexicon-based approach has an accuracy value of 70.3% while using a lexicon-based system improves the accuracy to 90.5%.