<![CDATA[Bogor Regency has significant tourism potential; however, available tourism information is mostly static and does not support preference-based recommendations. This study aims to develop a tourism recommendation system for Bogor Regency using a content-based filtering approach by integrating N-Gram, TF-IDF, and IndoBERT methods. The tourism destination dataset was collected through web scraping from online tourism sources and processed using text preprocessing techniques. Feature extraction was performed using N-Gram and TF-IDF to capture lexical similarity, while IndoBERT was trained using an Unsupervised SimCSE approach to generate contextual semantic representations. Destination similarity was calculated using cosine similarity, and system performance was evaluated using Precision, Recall, and F1-Score under Top-3, Top-5, and Top-10 scenarios. The experimental results show that the N-Gram and TF-IDF approach achieved the highest Precision of 63.95% in the Top-3 scenario and an F1-Score of 16.71% in the Top-10 scenario, indicating strong category consistency. Meanwhile, IndoBERT provided more context-aware recommendations with lower Precision, demonstrating its ability to capture semantic similarity beyond keyword matching. These findings indicate that lexical and semantic approaches complement each other and can be effectively combined to support more flexible and adaptive tourism recommendation systems.]]>
Copyrights © 2026
