Revolutionizing the Future of Automated Subjective Answer Sheet Evaluation System with Machine Learning and LLMs
Balla Aarathisree, Sujit Sarkar, Bhanu Durgesh Dammala, Mrityunjoy Panday, Neha Sharma · 2024
Human evaluation of handwritten answer sheets is very inefficient, inaccurate, and potentially biased. This research work serves the purpose of addressing these drawbacks by developing an automated grading system that leverages advanced OCR technologies and machine learning algorithms. The system, utilizes Google Cloud Platform (Cloud Vision API) and Gemini 1.5 Pro for Optical Character Recognition, to process large capacity of answer sheets with variation in handwriting styles more accurately than the traditional methods. This leads to a significant increase in text extraction accuracy and grading reliability. Therefore, as comparison with the existing solutions demonstrates, the new system has an increased level of accuracy and efficiency, reducing human error and fairly assessing student performances. Thus, implementation of changes in traditional grading system may influence the feasibility of this work, giving educators the opportunity to engage more time in instructional activities.