SMART GRADER USING ADVANCEDNATURAL LANGUAGE PROCESSING FOR ACADEMIC ASSESSMENTS

Authors

  • P. Rakesh,Dr. G. Narsimha Author

DOI:

https://doi.org/10.62643/

Abstract

Manual marking of descriptive handwritten tasks can be time-consuming and can cause inconsistent marking and late marking. This study presents an AI Assignment Grader which combines Optical Character Recognition, Natural Language Processing, semantic similarity, concept coverage, plagiarism analysis, automated grading, as well as AI-based feedback and report production in a role-based online application. Teachers design task and develop sample responses. Student answers are handwritten into pictures, PDFs or DOCX documents. PaddleOCR is a machine reading tool for uploading documents to obtain machine-readable texts. Using Sentence Transformer embeddings and cosine similarity, student responses are compared to reference answer . Marks, grades and notification of plagiarism are based on the concept coverage and similarity of submissions. The online application is powered by Flask, the academic data is stored in SQLite, PyMuPDF and Python-docx are used for the uploaded documents, and ReportLab is used to create downloadable academic evaluation reports. The project does not rely on a set dataset to learn a new classification model, rather the actual inputs are the instructor's answer key and student's assignments that are provided dynamically. Thus the implementation is evaluated based on the successful running of OCR, semantic analysis, plagiarism analysis, grading, feedback, analytics and report generation. The approach devised lowers the repetitious evaluation work, yet gives structured results and keeps the instructor in charge of the ultimate academic judgments. Keywords— Automated grading, handwritten assignments, PaddleOCR, Sentence Transformers, cosine similarity, plagiarism detection, Flask.

Downloads

Published

31-07-2026

How to Cite

SMART GRADER USING ADVANCEDNATURAL LANGUAGE PROCESSING FOR ACADEMIC ASSESSMENTS. (2026). International Journal of Engineering Research and Science & Technology, 22(3(1), 1084-1090. https://doi.org/10.62643/