Νικόλαος Λυμπιτάκης, Κατασκευή ενός RAG Chatbot για
Ξενοδοχειακή Υποστήριξη
vi
Abstract
This undergraduate thesis addresses the design and implementation of an intelligent
chatbot based on the Retrieval-Augmented Generation (RAG) architecture, applied to the
hospitality sector.
The motivation behind the development of this system lies in the significant
limitations of traditional human customer service, such as high operational costs, restricted
availability hours, and inconsistency in service quality. The RAG architecture was selected
as the most suitable approach, as it combines the advantages of Large Language Models
(LLMs) with information retrieval mechanisms from external knowledge bases, thereby
overcoming the static knowledge constraints of conventional models.
The implementation leverages modern technologies including: Python as the
primary programming language, LangChain for managing the data flow and interfacing with
the DeepSeek language model, the FAISS library for storing vector representations
(embeddings) and enabling fast semantic search, Streamlit for building a user-friendly
graphical interface, and SQLite as a lightweight database for hotel information and
reservations.
The system was designed according to specific requirements and use cases, and
evaluated through a structured methodology using quantitative metrics. The final chatbot is
capable of answering questions related to hotel amenities, policies, and room availability,
in both Greek and English.
The results confirm the effectiveness of the RAG architecture as a complete solution
for hospitality applications, delivering direct reductions in operational costs and an
improved user experience. The thesis concludes with a SWOT analysis and proposals for
future extensions, laying the groundwork for further development of intelligent customer
service systems in the tourism sector.
Keywords
RAG, Chatbot, Large Language Models, Artificial Intelligence