Sentiment Analysis and Topic Modeling of Tourist Attractions in Gresik Regency Using the BERT Method

Abstract

_{Tourism is one of the key sectors in national economic growth as well as a pillar of regional community welfare. Gresik Regency in East Java has considerable tourism potential, with more than 100 destinations covering religious tourism, natural attractions, and family recreation. However, tourist visit data from 2022–2024 shows a declining trend that requires an in-depth evaluation of visitor perceptions and experiences. This study aims to analyze public sentiment toward tourist destinations in Gresik Regency and identify the main topics of concern for tourists. The research data was collected from Twitter and Google Maps within the period of 2021–2024 using crawling techniques. Sentiment analysis was carried out with IndoBERT, while topic modeling was conducted using BERT. The results indicate that tourism reviews are dominated by positive sentiments highlighting the uniqueness of religious destinations, natural beauty, and family recreation atmosphere. However, negative sentiments were also found, emphasizing issues related to facilities, cleanliness, staff services, and accessibility to the sites. Topic modeling successfully grouped tourist opinions into coherent themes, and evaluation with coherence scores demonstrated good quality outcomes. The study concludes that although Gresik has strong tourism appeal, challenges in facility management, services, and digital promotion need to be addressed immediately. The integration of sentiment analysis and BERT-based topic modeling has proven effective in providing comprehensive insights into tourist perceptions and can serve as a basis for formulating regional tourism development strategies.}