AN OCR Based Intelligent System for Prediction of Marks from Hardcopies

Volume 20, Issue 1,  2026

Download

Author(s):

Jamil Ahmed* Department of Computer Science, SZABIST University, Larkana Campus, Pakistan, jamil.chandio@gmail.com

Mohammad Muzamil Department of Computer Science, SZABIST University, Larkana Campus, Pakistan, soomromohammadmuzamil@gmail.com

Shahzaib Shah Department of Computer Science, SZABIST University, Larkana Campus, Pakistan, syed669shahzaib@gmail.com

Hassan Ali Department of Computer Science, SZABIST University, Larkana Campus, Pakistan, rajahassanali25@gmail.com

Abstract The manual assessment of student answer papers (MASAC) is a critical constraint in education, characterized by subjectivity, inconsistency, and major time delays. This paper presents the design and implementation of an AI-powered automated marks assessment system (PAMAS) that uses cutting- edge deep learning models to create an end-to-end grading pipeline. The system implements LightOnOCR (optical character recognition) for text extraction from images of handwritten (IoH) and typed answer sheets (AS), effectively converting visual data into machine-readable text. The extracted answers are then evaluated by the DeepSeek-R1 large language model, which performs a sophisticated semantic analysis against a provided model answer and marking scheme. A key innovation of our system is its dual-output capability: it generates both an accurate numerical score and detailed, personalized feedback for the student. Experimental results on a dataset of 150 short-answer questions demonstrate a scoring accuracy of 91.2% compared to human expert graders, with a dramatic reduction in grading time from hours to seconds per script. The system proves the viability of integrating specialized OCR and advanced LLMs to create a scalable, objective, and efficient assessment tool that benefits both educators and learners.
Keywords automated Assessment, optical character recognition (OCR), large language model (LLM), educational technology
Year 2026
Volume 20
Issue 1
Type Research paper, manuscript, article
Journal Name Journal of Information & Communication Technology
Publisher Name ILMA University
Jel Classification -
DOI -
ISSN no (E, Electronic) 2075-7239
ISSN no (P, Print) 2415-0169
Country Pakistan
City Karachi
Institution Type University
Journal Type Open Access
Manuscript Processing Blind Peer Reviewed
Format PDF
Paper Link https://jict.ilmauniversity.edu.pk/journal/jict/20.1/3.pdf
Page 22-27