Evaluation of student's handwritten answers to Arabic poetry by an automated system
Huda Sabah Shuker, Raheem Ogla, Abdul Monem S. Rahma · IET conference proceedings. · 2025
This paper presents a novel deep learning (DL)-based model and system for automatic assessment of Arabic handwriting with a focus on student responses to Arabic poetry. To overcome the difficulties involved with cursive writing and diacritics of Arabic script, the proposed methodology combines the Deep Convolutional Neural Network (DCNN) with the Bidirectional Long Short-Term Memory (BiLSTM) networks and Connectionist Temporal Classification (CTC) loss. The training and testing were done using Kaggle’s Arabic Handwritten Characters to ensure that the system tested came with various styles of handwriting. These performance measures were achieved through tested whereby the system delivered an astonishing 97. 20% accuracy, 97. 10% precision, and 97. 30% recall can be achieved using the proposed methods as compared to the existing techniques. The set of preprocessing involving normalization and binarization has helped in enhancing the text legibility, which augments the pre-existing prowess of the system in character recognition and sequence modelling. The ability to process information in real time guarantees swift provisioning of results and feedback and places the system in a good standing for its uses in the educational environment as an automatic marker, handwriting analyzing system, as well as a personalized tutoring system. This study not only contributed to the field of Arabic handwriting recognition but also provided the opportunities for further developments in automated handwriting assessment systems.