Semantic Refined Prompting based Automated Essay Scoring System
Vishvasundar Senthilnathan, Sakthi Vaibhav M, Alexander R · 2025
The usage of Large Language Models (LLMs) has made significant contributions in various areas including essay scoring. However, its performance is not comparable to human grades due to the reasons that the auto-generated scores are not done based on semantic understanding. Existing research on this topic focuses on grading with predefined rubrics. Since grading can’t be relied only on the predefined rubrics in this paper, we propose an LLM-based scoring mechanism that considers the predefined and semantic refined rubrics. Additionally, the scores obtained are evaluated for consistency and fairness. This system is evaluated using the ASAP dataset and the outcomes demonstrate the effectiveness of the approach in terms of RMSE and MAP.