Leveraging LangChain for Enhanced Tourism Guidance: A Retrieval-Augmented Generation Approach for SmartTour Chatbot

Abstract

SmartTour chatbot is designed to provide accurate and relevant tourism guidance to travelers visiting Barru Regency. Developed using the Streamlit framework, the application offers a user-friendly interface where users can interact with the chatbot to receive information about local attractions, cultural heritage, and tourism-related services. The chatbot uses GPT-4.1 and leverages a Retrieval-Augmented Generation (RAG) approach, integrating contextual data extracted directly from a tourism guide PDF into a vector database to ensure the accuracy of responses. Text preprocessing, including text cleaning and tokenization, is implemented to enhance the system's ability to process and understand user queries effectively. The system's performance was optimized with parameters such as chunk_size = 1500, chunk_overlap = 150, and k = 9 to improve data retrieval efficiency and ensure the relevance of responses. The system was evaluated with 10 valid tourism-related questions designed to assess the chatbot's accuracy in providing relevant answers. The performance was tested under two conditions: with and without text preprocessing, achieving an accuracy rate of 80% with preprocessing and 60% without. This study demonstrates the effectiveness of combining large language models with retrieval systems to create a dynamic and reliable tourism assistant, offering valuable insights into improving tourism services in Barru Regency and similar regions.