Exécution de notre outil de synthèse sur un nouvel article
Vérifiez l’article ci-dessous, puis lancez sa synthèse.
A Course-Scoped RAG Domain Model for an AI Teacher Assistant: Architecture and Pilot Evaluation
Battsetseg Erdenebat (Mongolian University of Science and Technology), Uranchimeg Tudevdagva (Mongolian University of Science and Technology), Amarzaya Amartuvshin (National University of Mongolia), N. I. Lomakin (Volgograd State Technical University), Ilhom Boynazarov (Tashkent University of Information Technology)
Résumé : Large Language Models (LLMs) are increasingly used in higher education, yet general-purpose systems may generate unsupported or overly general answers when they are not grounded in verified course materials. This paper presents a course-scoped Retrieval-Augmented Generation (RAG) Domain Model for an AI Teacher assistant as a foundational component for a future Intelligent Tutoring System (ITS). The prototype uses course-approved materials as a retrieval scope and returns answers with file and page references through a Telegram-based interface. A pilot implementation was evaluated with a user survey$\boldsymbol{(} \mathbf{N}=\mathbf{3 3} \boldsymbol{)}$, representative source-cited interactions, and a structured plan for subsequent source support evaluation. The survey indicates positive perceived usefulness and reuse intention, while response speed remains a limitation. The contribution is an architecture and pilot evaluation framework for traceable, course-bounded AI assistance rather than a complete ITS or a randomized learning-outcome study.