MathScribe: An AI-Powered Framework for Handwritten Mathematical Expression Recognition and Solving

  • Unique Paper ID: 177236
  • Volume: 11
  • Issue: 12
  • PageNo: 331-340
  • Abstract:
  • Handwritten mathematical expression recognition and solving remain critical challenges in educational technology and accessibility, necessitating accurate, real-time solutions. This paper introduces MathScribe, a hybrid deep learning framework that combines convolutional neural networks (CNNs) for handwritten character recognition with a symbolic computation engine to parse and solve mathematical expressions, including differentiation and integration. Our system employs a dual-model CNN architecture (VGG16 and ResNet) to extract spatial and hierarchical features from handwritten digits and operators, fused with structural expression parsing for robust interpretation. The processed input is then evaluated using SymPy-based symbolic computation, providing step-by-step solutions. Deployed via a Streamlit web application, MathScribe allows users to upload images or draw expressions in real time, delivering solutions with an accuracy of 98.7% on the MNIST dataset and 94.2% on custom operator benchmarks. Experimental results demonstrate strong performance, with an F1-score of 97.3% and near-perfect ROC-AUC (0.998) for expression recognition. The minimal training-validation loss gap (0.15) confirms generalization efficacy. MathScribe bridges the gap between raw handwritten input and advanced mathematical problem-solving, offering a scalable tool for education and accessibility.

Cite This Article

  • ISSN: 2349-6002
  • Volume: 11
  • Issue: 12
  • PageNo: 331-340

MathScribe: An AI-Powered Framework for Handwritten Mathematical Expression Recognition and Solving

Related Articles