Requirement Alalysis of the AISE Platform
The requirement analysis of the AISE platform was used to establish the architectural and functional foundation for the AI-based student skill assessment system which covered the need to support multilingual input (Greek, Serbian, Croatian) from schools in four countries. Based on this analysis, a Retrieval-Augmented Generation (RAG) architecture was adopted which combined Large Language Models (LLMs) as the core technical approach for automated essay assessment, thereby replacing the originally envisaged conventional machine learning pipeline with a more advanced and pedagogically interpretable solution. The requirements specified three independent, dimension-specific grading axes (or dimensions), namely the Content, Organization, and Language, where each one scored on a four-level qualitative scale (Excellent, Very Good, Good, Marginal) to reflect standard human grading rubrics. The analysis also defined the system’s deployment requirements, including a FastAPI-based service architecture intended for integration with the AISE platform and hosting on a dedicated university server to ensure scalability and accessibility for the participating schools. These requirements directly informed the design of the dataset preparation pipeline and AI modelling activities carried out in the subsequent stages of WP2.