MESA Model for Evaluating Subjective Answers Using AI

Manual evaluation of subjective answers in academic settings is time-consuming and prone to inconsistencies and biases. This paper presents a Model for Evaluation of Subjective Answer (MESA), an AI/ML-based system that automates the assessment of handwritten subjective responses. It combines Optical Character Recognition (OCR) and advanced Natural Language Processing (NLP) techniques such as MPNet, Paraphrase models, and GPT-4.0 Mini API to evaluate answers semantically. The system ensures faster, fairer, and scalable evaluation while detecting potential plagiarism. Experimental results demonstrate significant accuracy and performance in subjective grading automation. The reason behind doing so is that it tests the analytical as well as the reasoning ability of a student while also being extremely accurate and efficient. Hence there is a need for a system that optimizes the task of repetitive answer sheet corrections and provides optimal accuracy. The proposed model has come up with a reliable system based on work that formulates the marks scored by the student based on the sentence similarity, Jaccard similarity and grammar of the model answer and student answer.