Automatic Grading of Answer Sheets using Machine Learning Techniques

Automating the grading process for question-answer sheets represents a significant challenge, particularly when dealing with traditional hard copy papers.This initiative aims to reduce the time and expenses associated with manual grading, a task that typically consumes 2-3 days for teachers to complete.Leveraging advanced Natural Language Processing (NLP) and Machine Learning (ML) methodologies, including XGBoost, Ridge Regression, and Naive Bayes, we have developed a system for automatic grading using prepossessed OCR datasets.This system learns from a historical dataset of student question-answers, with a focus on two primary objectives: scoring short-answer questions and providing constructive feedback to students.Additionally, we assess the performance and accuracy of the system using standard evaluation techniques such as Precision, Recall, and F-measure.Our experimental results demonstrate an impressive 89% accuracy in grading student answer sheets.