CHIMLE: An LLM-based chatbot for managing student academic databases

Abstract

This paper presents CHIMLE, an intelligent chatbot prototype based on a Large Language Model (LLM) designed to interact directly with institutional student academic databases. The proposed system supports natural language-based access to academic information, automated and personalized report generation, responses to frequently asked questions, and assistance with administrative processes. Its objective is to enhance accessibility, usability, and operational efficiency in academic information management within higher education institutions. The solution is implemented in Google Colaboratory integrating a generative language model with SQL-based tools that allow secure and dynamic interaction with structured institutional data. The architecture supports text-to-SQL translation, multi-turn conversational queries, and real-time data retrieval. The prototype was evaluated through a set of representative academic queries related to student performance, course information, and aggregated statistics. The results demonstrate that CHIMLE can accurately interpret user intentions and generate reliable responses with low execution time, highlighting its potential as a flexible and scalable support tool for academic environments. This work extends a previously reviewed preliminary version by incorporating refined architecture, a complete implementation, additional functional tests, and an expanded discussion in relation to recent advances in LLM-based educational chatbots.