2026· International journal of research and innovation in applied science· Vol 11, pp. 423-441· 0 citations
TL;DR
Although exact word-level recognition remained difficult because of the extremely limited dataset size, the proposed framework successfully learned meaningful sequential patterns and produced increasingly structured Kirundi-like predictions.
Abstract
Handwritten Text Recognition (HTR) has experienced remarkable progress with the development of deep learning techniques. However, most existing studies focus on high-resource languages for which large annotated datasets are readily available. In contrast, low-resource languages remain largely underrepresented in handwriting recognition research due to the scarcity of handwritten corpora, linguistic resources, and benchmark datasets. This paper presents a handwritten word recognition framework for Kirundi, a low-resource Bantu language spoken primarily in Burundi. The proposed system employs a Convolutional Recurrent Neural Network (CRNN) combined with Connectionist Temporal Classification (CTC) for end-to-end sequence recognition without explicit character segmentation.
To address severe data scarcity, a small handwritten Kirundi dataset consisting of manually collected word samples was constructed and annotated. Data augmentation techniques, including rotation, translation, Gaussian noise, Gaussian blur, and elastic distortion, were applied to increase sample diversity. In addition, synthetic handwritten-style data were generated to further expand the training set. Three experimental configurations were investigated: real handwritten data only, real data with augmentation, and real data combined with augmentation and synthetic handwritten-style images.
Experimental results demonstrate that synthetic data generation improved recognition performance and reduced Character Error Rate (CER) from 0.8354 to 0.7560, corresponding to an approximate relative improvement of 9.5%. Although exact word-level recognition remained difficult because of the extremely limited dataset size, the proposed framework successfully learned meaningful sequential patterns and produced increasingly structured Kirundi-like predictions. The study establishes an initial benchmark for Kirundi handwritten word recognition and highlights the potential of synthetic data generation for low-resource handwriting recognition tasks.
Offline handwritten word recognition has progressed from whole-word classification to sequence transcription, yet many systems depend on large annotated corpora and exploit lexical regularities over explicit character evidence. This paper presents an alternative formulation for Arabic offline handwritten word recognition, treating characters as spatial objects detected via a Faster Region-Based Convolutional Neural Network rather than symbols generated by a one-dimensional decoder. We construct and release a character-level annotated subset of 2153 handwritten word images from a standard Arabic benchmark, exporting matched detection, sequence, and word-class labels. We also introduce an open-source subword exchange toolkit that creates a controlled structural-generalization benchmark by swapping subwords while preserving handwriting style. Experiments compare the proposed detector against whole-word and sequence-based baselines on both the original held-out split and the perturbed benchmark. Results show sequence models degrade sharply under structural recombination, whereas the proposed detector remains stable, achieving a 26.56% character error rate and 70.0% word accuracy on the perturbed benchmark. These findings demonstrate that explicit character localization provides a robust, data-efficient alternative for Arabic handwritten text recognition in low-resource settings.
Sofiane Medjram, Ruwaidah Saud Alnejaidi· Applied Sciences· 0 citations
This is the first study to introduce both a real handwritten Urdu word dataset and a diffusion-generated synthetic dataset, and develops a unified word recognition model trained jointly on handwritten and printed Urdu word data, leading to improved recognition robustness and performance.
Wahid Hussain, Shahbaz Hassan, I. Hassan et al.· IEEE Access· 0 citations
The digitisation of historical, administrative, and personal documents in Bulgarian faces considerable challenges due to the lack of robust Optical Character Recognition (OCR) and Handwritten Text Recognition (HTR) systems tailored for the Cyrillic alphabet. While modern Vision-Language Models (VLMs) and large transformer-based architectures achieve state-ofthe-art results, their performance and resource efficiency on low-resource languages remain prohibitive for decentralised, privacy-preserving applications. In this paper, we present a comprehensive, resource-efficient pipeline for extracting printed and handwritten Cyrillic text. Our system integrates advanced image preprocessing, YOLO-based document structure and table recognition, and a Permuted Autoregressive Sequence (PARSeq) model trained specifically for Bulgarian. We generated custom synthetic Bulgarian cursive datasets to mitigate the severe lack of real-world training data. Our evaluation indicates that the specialised PARSeq model outperforms traditional OCR tools such as Tesseract and EasyOCR on our custom degraded printed test set, and provides a practical, resource-efficient baseline for handwriting recognition compared to a modern local VLM (Qwen3-VL-4B). Finally, we discuss the discrepancy between synthetic and real handwritten data, highlighting the urgent need for a standardised, annotated Bulgarian HTR dataset.
D. Halachev, Ivan Koychev· Computational Linguistics in...· 0 citations
To overcome the drawbacks of current techniques that usually concentrate on a single category and to tackle the difficult problem of identifying heterogeneous handwritten text in real-world situations like exam papers. A Handwritten Letter Classification Module therefore a Handwritten Text Intelligent Recognition Module are the two main parts of the system, and they operate together. The recognition module improves recognition accuracy by dynamically choosing specialised sub-networks for each category based on these findings. A Context-aware Recognition Optimisation Module is presented to further reduce mistakes resulting from identical character forms and different handwriting styles. This work creates a diverse, integrated handwritten text collection in recognition of the shortcomings of current public handwriting datasets, especially their lack of variation in character types and writing styles. The real-world applications, the dataset incorporates samples from several sources, such as English letters, numbers, and mathematical symbols. The proposed system employed the algorithm ConvNeXt-Tiny with maximum accuracy of 99.5 percentage.
K. Raghavi, A. Leena, M. Robinson· International Conference Com...· 0 citations
An edge-aware line-level HTR framework that extends a CNN-Transformer baseline with a learnable edge-extraction channel and Squeeze-and-Excitation channel attention and shows that combining learnable structural cues with channel-wise attention has improved robustness for degradation-prone historical manuscript collections.
Bilal Abdulrahman, Farhan Mohamed· Journal of Human Centered Te...· 0 citations
A deep learning-based handwritten character recognition system that leverages Convolutional Neural Networks for automatic feature extraction and classification and highlights the effectiveness of deep learning techniques in enhancing recognition performance and reducing classification errors compared to conventional machine learning methods.
Shwetha M R Shwetha M R, Kowshik S S Kowshik S S· International Scientific Jou...· 0 citations