JOBCHECKER: AUTOMATING ASSESSMENT OF HANDWRITTEN STUDENT WORK

Abstract

In educational institutions, the grading of handwritten student assignments still seems to be a time-consuming and prone-to-error process. Designed to improve grading efficiency, consistency, and accessibility for both teachers and students, JobChecker is an AI-powered tool meant to automate the evaluation of written student work. The system uses OpenAI’s GPT-4 model to assess and generate comments on handwritten input and incorporates Optical Character Recognition (OCR) technology to transform it into digital form.

A key aspect that distinguishes JobChecker is its holistic integration of advanced OCR for handwritten text with the sophisticated contextual understanding of the GPT-4 model, enabling accurate assessment directly from handwritten submissions—a capability not commonly found in existing automated grading solutions.

JobChecker supports multiple subjects and sections, provides multilingual feedback, and is accessible via web and mobile applications. The frontend and mobile interfaces use ReactJS and SwiftUI respectively for a responsive user experience; its backend architecture is driven by Java Spring Boot and PostgreSQL. The project includes a comparative evaluation of several state-of-the-art language models, highlighting GPT-4’s better contextual understanding and API accessibility.

Along with an analysis of technical difficulties faced during implementation, this report covers the design, development, and performance evaluation of JobChecker. By reducing the manual workload on teachers and providing instant feedback to students, JobChecker shows how artificial intelligence might modify educational processes.