Scoring System for Academic Essay Questions with Personalised Feedback Generation

This study presents an advanced Automated Essay Scoring (AES) system that seeks to enhance academic evaluation systems by integrating an LLM such as BERT, GPT-3, and GPT-4. This study employs few-shot learning technology to enable models to learn to adapt to different themes of essays and student styles. Provides human-in-the-loop feedback to achieve accuracy and scalability by incorporating manual annotation and LLM-assisted annotation. The combination of automated scoring and actionable trait-specific feedback integrates this broad-generating scoring system with individualized pedagogical support. Evaluation indicates that this system would likely lead to rather outstanding educational improvements through providing precise constructive alignments to academic standards and rubric space. This work extends AES technologies for personalized learning against existing gaps of feedback quality and accessibility.