Building a Bogor Tourism Recommendation System with N-Gram, TF-IDF, and IndoBERT

Bogor offers many destination options, but too many choices can make users unsure where to start. Someone may want waterfalls, family-friendly places, educational tourism, or destinations for a short trip, but they may not always know the right keywords.

BogorXplore was created to make exploration simpler. Users can browse a destination catalog, search by interest, filter categories, open tourism details, view locations on a map, and get recommendations for similar places in one practical flow.

The goal of this project is to help users make decisions faster: which places are interesting, where they are located, what facilities they offer, and which other destinations are similar.

BogorXplore's main strength is a more relevant search experience. Users are not only given a static list, but are guided toward destinations that match the context of their needs.

The project also combines tourism information with maps. Users do not stop at reading descriptions; they can understand where a place is located and start planning their route.

For a more stable experience, the application still includes fallbacks. If the machine learning service is inactive, basic search and recommendations can still work so the product does not immediately feel broken.

Behind the simple experience, BogorXplore uses a natural language processing approach. Destination data is processed so the text becomes cleaner and ready for comparison, from character cleaning, tokenization, and stopword removal to Indonesian stemming.

N-Gram and TF-IDF are used to understand similarity between destination descriptions. For example, places related to waterfalls, nature, or family trips can be found as similar destinations even when their names are different.

IndoBERT is used for semantic search, allowing the system to understand user queries more flexibly. This makes the search experience feel smarter than basic keyword matching.