RAG-based architecture enhanced specialized chatbot generation
Δημιουργία εξειδικευμένου chatbot βασισμένο στην αρχιτεκτονική RAG

Master Thesis
Author
Ntourmas, Ioannis
Ντούρμας, Ιωάννης
Date
2025-03View/ Open
Keywords
Retrieval-Augmented Generation (RAG) ; Natural Language Processing (NLP) ; Chatbot architecture ; Domain-specific dialogue systems ; Knowledge-enhanced generationAbstract
The growing need for specialized chatbots, tailored to distinct domains, is increasingly evident
across a wide range of industries and academic institutions. This thesis explores the design
and implementation of a chatbot grounded in a Retrieval-Augmented Generation (RAG)
architecture, specifically developed for the University of Piraeus. By integrating GPT-3.5 with a
data retrieval system based on MongoDB, the chatbot delivers accurate and context-relevant
answers for university-related queries. The dissertation focuses on the design principles
employed in Flask for backend development, the React-based frontend for a user-friendly
interface, and the optimization strategies that facilitate seamless data management.
Experimental evaluations confirm the chatbot’s effectiveness in providing domain-specific
answers, underscoring its flexibility and reliability for academic needs.