The hospitality industry relies significantly on online reviews as a key source of information that influences potential guests' decisions. However, conducting sentiment analysis on hotel reviews can be challenging due to the complexity of language and the diversity of contexts, particularly in Indonesian. This research aims to develop an optimized RoBERTa-based sentiment analysis model to enhance the accuracy of sentiment classification in Indonesian hotel reviews. It focuses on key factors such as facilities, cleanliness, location, price, and service. The methodology involves data collection through web scraping from the Traveloka platform, manual labeling, and text preprocessing. The RoBERTa model was trained and optimized through fine-tuning techniques and evaluated using metrics like accuracy, precision, recall, F1-score, and AUC. The results demonstrate that the optimized RoBERTa model achieves competitive performance; however, the IndoBERT model with Bayesian Optimization outperforms it, particularly in accuracy and efficiency in identifying positive and negative sentiments. This research is expected to significantly contribute to the development of a more effective and accurate aspect-based sentiment analysis (ABSA) for Indonesian hotel reviews. Additionally, it opens opportunities for applying NLP technology in the hospitality industry, which could also be extended to other review platforms, thereby improving sentiment analysis quality and assisting hotel managers in enhancing customer service and experience.
Kata Kunci: Aspect Based Sentiment Analysis; RoBERTa; IndoBERT; Fine-Tuning; Web Scraping