DEVELOPMENT OF AN AI CHATBOT MODEL FOR STUDENT ADVISING AT A CENTER FOR FOREIGN LANGUAGE AND INFORMATICS

Abstract

At present, chatbots are widely adopted across various sectors, particularly in educational institutions. Nevertheless, those relying on large language models often demand high-performance computing resources, which pose cost challenges for many local education centers. This study introduces an AI-based student advising chatbot that has been designed and piloted at the Center for Foreign Languages and Informatics, Can Tho University of Technology. The system aims to deliver academic advising services under limited-resource conditions, employing the Rasa framework in combination with LaBSE and PostgreSQL. A distinctive contribution of this work lies in building a CPU pipeline that does not depend on GPUs or commercial APIs, together with a detailed preprocessing procedure for Vietnamese text. The dataset includes 1,712 questions collected from students, advisors, and the Center’s Facebook page, categorized into 56 intents and 55 dialogue scenarios. Ten-fold cross-validation results indicate that the chatbot achieves 98.1% accuracy for intent classification and 99.9% for entity recognition. Experimental deployment further validates the feasibility of the proposed model in providing student advising services at educational institutions with constrained infrastructure.