AI-Driven OCR-Based Script Grading

Abstract

AI-Driven OCR-Based Script Grading System is a novel web-based solution designed to automate the evaluation of descriptive answers in educational assessments, addressing significant challenges associated with manual grading processes, such as inconsistent scoring and high time consumption. This system integrates cutting-edge technologies like Optical Character Recognition (OCR), Natural Language Processing (NLP), and advanced summarization techniques to streamline answer script evaluations effectively. EasyOCR is utilized to accurately extract textual content from scanned handwritten or printed student submissions. Subsequently, this extracted text undergoes summarization using the Hugging Face BART model to retain essential points and remove verbosity, facilitating precise comparative analysis. The summarized student responses are compared against predefined teacher answers through SequenceMatcher, calculating a semantic similarity score. This similarity score directly translates into objective grading, significantly reducing educator workload and eliminating potential grading biases. The Flask-based web platform provides a secure, intuitive, and efficient interface for educators, ensuring seamless integration into existing educational workflows. This automated grading system promotes consistency, enhances accuracy, and markedly improves turnaround times, ultimately enabling educators to dedicate greater attention toward instructional quality and student engagement.

Introduction

The evaluation of descriptive answer scripts in education is traditionally manual, time-consuming, and prone to subjective inconsistencies, human error, and fatigue. To address these issues, the paper proposes an AI-Driven OCR-Based Script Grading System that automates grading by combining Optical Character Recognition (OCR) and Natural Language Processing (NLP) technologies.

The system uses EasyOCR to extract text from handwritten or printed answer scripts, then employs the Hugging Face BART model to summarize responses. A semantic similarity algorithm, SequenceMatcher, compares student answers with teacher-provided model answers to generate objective similarity scores, which translate into marks. This approach enhances grading accuracy, consistency, and efficiency while reducing educators’ workload.

Built on a Flask web framework, the platform supports various input formats, customizable evaluation criteria, and generates detailed CSV reports to aid teacher assessment. The system streamlines the grading process, enabling timely and fair feedback.

The paper also reviews related AI-based grading tools and OCR technologies, highlighting their benefits and limitations, and presents a modular system architecture with clear workflows, sequence models, and use case diagrams. Testing demonstrated the system’s robustness across handwriting styles, image qualities, and its ability to deliver reliable, quick, and automated evaluations.

Conclusion

The proposed AI-Driven OCR-Based Script Grading System aims to revolutionize the evaluation of descriptive answers in educational assessments. This system combines advanced artificial intelligence technologies, primarily Optical Character Recognition (OCR) and Natural Language Processing (NLP), to automate and enhance the grading process effectively.

References

[1] Tobler, Samuel. \"Smart grading: A generative AI-based tool for knowledge-grounded answer evaluation in educational assessments.\" MethodsX 12 (2024): 102531 [2] Prerana M. S., Sagarika M. Chavan, Ramit Bathula, Sreenath Saikumar, and Geetha Dayalan. \"Eval.\" 2023 International Journal on Advanced Research, ISSN: 2147-67992, January 4, 2023. [3] Mahmud, Saikat, et al. \"Automatic multiple choice question evaluation using tesseract ocr and yolov8.\" 2024 IEEE Conference on Artificial Intelligence (CAI). IEEE, 2024. [4] B. K. Pattanayak, A. K. Biswal, S. R. Laha, S. Pattnaik, B. B. Dash, and S. S. Patra, \"A Novel Technique for Handwritten Text Recognition Using Easy OCR,\" Proceedings of the International Conference on Self Sustainable Artificial Intelligence Systems (ICSSAS 2023), pp. 1115–1118, Dec. 2023, DOI: 10.1109/ICSSAS57918.2023.10331704.

Copyright

Copyright © 2025 S Karuna Kumari, Sameer Razak B, Shaguftha Subedar, Sirish Gupta S, Dr. CN Sharif. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.