SmartPen: Online System for Automated Grading of Handwritten Answers using Deep Learning

Abstract

This paper introduces SmartPen, an intelligent, web-based system that automates the evaluation of handwritten answers using Optical Character Recognition (OCR) and semantic similarity models. It integrates Tesseract OCR for converting scanned scripts into text and leverages Sentence-BERT for context-aware semantic comparison of student responses with model answers. SmartPen provides educators with an interactive Streamlit-based interface to upload question papers, extract or input model answers, and evaluate student submissions using cosine similarity scoring. The system supports real time evaluation, manual verification of OCR output, and downloadable scoring summaries. Through AI-driven semantic understanding, SmartPen offers an accurate, scalable, and unbiased approach to assessing subjective answer scripts in academic environments.

Introduction

Problem:
Manual grading of handwritten student answers is time-consuming, subjective, and not scalable as class sizes increase.

Solution:
SmartPen is a voice-assisted, intelligent, web-based platform that automates grading by combining Optical Character Recognition (OCR) with semantic similarity analysis. It uses Tesseract OCR to extract text from handwritten scripts and Sentence-BERT to semantically compare student answers with model answers, enabling context-aware evaluation.

Objectives:

• Improve grading efficiency, accuracy, and fairness.

• Reduce manual effort and subjectivity.

• Provide real-time feedback and scalable assessment with an intuitive interface.

Literature Review:
Previous methods used keyword matching, NLP techniques, and machine learning but had limitations in capturing deep semantic meaning, scalability, or usability. Some systems relied on cloud services raising privacy concerns. Sensor-based handwriting quality assessment exists but does not address semantic content evaluation. Recent work integrating OCR and BERT shows promise but lacks real-time manual verification and user-friendly interfaces.

Proposed Methodology:

• Upload scanned answer sheets and questions via a Streamlit web interface.

• Extract text using Tesseract OCR.

• Educators input or verify model answers.

• Generate semantic embeddings for answers using Sentence-BERT.

• Compute cosine similarity for semantic scoring.

• Assign marks based on similarity scores.

• Provide detailed evaluation reports downloadable by educators.

System Architecture:
Consists of layers for web interface, input handling, OCR, semantic similarity, data management, scoring, and report generation.

Implementation & Results:

• Achieved over 75% OCR accuracy on diverse handwriting under good scan quality.

• Semantic scoring matched manual grading in 92% of test cases, demonstrating fairness and accuracy.

• Generates transparent, detailed reports to support evaluation and feedback.

Conclusion

This research demonstrates the effectiveness and practicality of SmartPen as an AI-driven solution for automating the evaluation of handwritten academic responses. By integrating OCR with semantic similarity analysis, SmartPen significantly reduces grading time, improves consistency, and minimizes human bias in subjective assessments. The system offers a streamlined and user-friendly interface for educators, ensuring both transparency and scalability in academic evaluation. The implementation validates that intelligent automation can enhance the accuracy and fairness of grading, making it a valuable tool for educational institutions seeking efficient and equitable assessment processes.

References

[1] N. Bharambe, P. Barhate, and P. Dhannawat, “Automatic Answer Evaluation Using Machine Learning,” International Journal of Information Technology (IJIT), vol.?7, no.?2, pp.?1–3, Mar.–Apr.?2021. Usha Mittal Institute of Technology, SNDT Women’s University, Maharashtra, India. http://www.ijitjournal.org/ [2] S. P. Raut, S. D. Chaudhari, V. B. Waghole, P. U. Jadhav, and A. B. Saste, “Automatic Evaluation of Descriptive Answers Using NLP and Machine Learning,” International Journal of Advanced Research in Science, Communication and Technology (IJARSCT), vol.?2, no.?1, pp.?735–745, Mar.?2022. Navsahyadri Institute of Technology, Pune, India. DOI: 10.48175/IJARSCT-3030 [3] Prerana M. S., Sagarika M. Chavan, Ramit Bathula, Sreenath Saikumar, and Dr. Geetha Dayalan, “Eval – Automatic Evaluation of Answer Scripts using Deep Learning and Natural Language Processing,” International Journal of Intelligent Systems and Applications in Engineering (IJISAE), vol. 11, no. 1, pp. 316–323, 2023. PES University, Bengaluru, India. https://ijisae.org/index.php/IJISAE/article/view/2541 [4] M. Bublin, F. Werner, A. Kerschbaumer, G. Korak, S. Geyer, L. Rettinger, E. Schönthaler, and M. Schmid-Kietreiber, “Handwriting Evaluation Using Deep Learning with SensoGrip,” Sensors, vol.?23, no.?11, pp.?5215, May 2023. University of Applied Sciences FH Campus Wien, Vienna, Austria. DOI: 10.3390/s23115215 [5] M. Kulkarni, G. Adhav, K. Wadile, R. Chavan, and V. Deshmukh, “Digital Handwritten Answer Sheet Evaluation System,” International Journal of Scientific Research in Computer Science, Engineering and Information Technology, Savitribai Phule Pune University, Pune, India, Feb. 2024. DOI: 10.21203/rs.3.rs-3978232/v1

Copyright

Copyright © 2025 Shirisha Dyavanapally, G. Narasimham. 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.