Dublin Core
Title
TOURISTBOT: AI-POWERED MULTILINGUAL TRAVEL COMPANION FOR SEAMLESS TRAVEL ASSISTANCE IN BOSNIA AND HERZEGOVINA
Abstract
Tourists visiting Bosnia and Herzegovina often encounter language barriers and difficulties in accessing relevant travel information in real-time, hindering their overall travel experience. Existing solutions are either limited in multilingual support or fail to deliver seamless, contextually appropriate assistance. This project represents a practical solution, namely, a TouristBot, an AI-powered, multilingual virtual travel assistant, designed to enhance the travel experience for visitors of Bosnia and Herzegovina. The system employs Natural Language Processing (NLP), and trained models, to process the user queries in English, and uses Google Cloud’s Translation API for enabling multilingual support, thus returning the answer in the user's original language. The backend, built in Python, using Flask framework, follows three-layered architecture, thus firstly searching for the answer directly from the custom dataset, then using pretrained models for prediction if the threshold for the first layer is not achieved, and finally mBert model is being used for deep semantic understanding if the first two layers fail. Also the frontend is built in JS’s framework, React-Native, for cross-platform functionality. That modular development methodology ensures higher reliability of the TouristBot, and eliminates the problem of the single point of failure. Ultimately, the outcome is a fully functional cross-platform mobile app, which aims to contribute to the tourism industry by offering an innovative, user-friendly solution that bridges language gaps and streamlines the travel experience in Bosnia and Herzegovina.
Keywords
Natural Language Processing (NLP), virtual travel assistant, multilingual chatbot, cross-platform app, pretrained models, tourism in Bosnia and Herzegovina