Integrating Rasa Framework and Retrieval-Augmented Fine-Tuning for an Educational Advising Chatbot
Hai-Nam Dinh-Duong, Tien-Julian Ho, Trong-Nguyen Pham, Duy-Hoang Tran · 2025
In this paper, we present a novel approach to building an advanced advising chatbot by combining the Rasa framework with Retrieval-Augmented Fine-Tuning (RAFT). Rasa provides robust capabilities for building custom conversational agents, while RAFT allows the chatbot to access and incorporate dynamic, domain-specific knowledge in real time. This addresses the limitations of static chatbots by enabling accurate, context-aware responses tailored to individual user needs. We present a case study developing a chatbot for the Vietnamese Standardized Test of English Proficiency (VSTEP) exam, demonstrating its ability to provide personalized guidance on registration, preparation, and result retrieval. We evaluate the proposed chatbot architecture based on response accuracy and scalability for educational advising. Our findings demonstrate the potential of integrating Rasa and RAFT to create intelligent, flexible, and efficient advising systems adaptable to various educational contexts.