AI Assisted System for Automated Evaluation of Entity-Relationship Diagram and Schema Diagram Using Large Language Models
Raji Ramachandran, Parvathy Vijayan, Athulya Anilkumar, Veena Gangadharan · Big Data and Cognitive Computing · 2025
Automated assessment in education has seen rapid growth with the integration of AI, particularly for objective and structured tasks. However, evaluating open-ended design problems such as Entity Relationship (ER) diagrams and relational schemas remains a significant challenge due to the variability in valid representations. This paper proposes an AI-assisted framework using Large Language Models(LLMs) to interpret natural language database scenarios, generate reference ER diagrams and schemas in PlantUML format. and compare student submissions against the system generated solutions to assess correctness. We proposed a novel scoring mechanism for evaluating the semantic and structural similarity of entities, relationships, keys, and table mappings, rather than relying on exact syntax matching. Additionally, manual verification of AI-generated reference outputs enables human oversight and refinement, making the system a supportive tool rather than a replacement for educators. This approach offers scalable, intelligent evaluation for database design tasks, reducing the manual grading effort while ensuring fair and concept-driven assessment. Experimental results demonstrate the system’s effectiveness in accurately evaluating varied student submissions while maintaining adaptability across different design styles.