kailabresearch/khld-kurdish-handwritten-sentences

KHLD: A Large-Scale Benchmark of the Kurdish Handwritten Lines Dataset for Low-Resource Central Kurdish (Sorani)

Dataset Summary

The Kurdish Handwritten Lines Dataset (KHLD) is a large-scale benchmark designed for handwritten text recognition (HTR) and optical character recognition (OCR) systems specifically for Central Kurdish (Sorani). It consists of 47,944 handwritten line images organized into 4,802 unique sentences, with 10 variations each.

The dataset captures the unique orthographic features of the Kurdish alphabet (based on Arabic script with modifications), including specific characters such as ڤ, ڕ, ڵ, ێ, and ۆ.

• Total Images: 47,944
• Unique Sentences: 4,802
• Variations per Sentence: 10
• Format: JPEG (Grayscale, resized)
• Resolution: 600–1,200 DPI (original scan)

Dataset Structure

The dataset is organized into folders where each folder corresponds to a unique Kurdish sentence. Each folder contains 10 image files representing different handwriting samples of that sentence.

Data Fields

• image: The handwritten line image.
• text: The ground truth typed Kurdish sentence (Sorani).
• folder_id: The ID of the sentence group.
• variation_id: The ID of the specific handwriting variation.

Dataset Creation

Curation Rationale

Kurdish (Sorani) is a low-resource language in the field of AI and OCR. This dataset was created to fill the gap in high-quality, large-scale benchmarks for Kurdish handwritten text recognition.

Source Data

Sentences were collected from a variety of sources including Kurdish websites, books, articles, and social media to ensure a mix of formal and colloquial language styles.

Data Collection and Processing

• Participants: Samples were collected from native Sorani speakers (ages 15–55) at universities and institutes in Erbil, Kurdistan Region, Iraq.
• Segmentation: Line segmentation was performed using a fine-tuned YOLOv8 model (achieving 95%+ accuracy) and OpenCV template matching.
• Preprocessing: Images were converted to JPEG, grayscaled, and resized for consistency.

Citation Information

If you use this dataset in your research, please cite:

@misc{ibrahim2026khld,
  title={KHLD: A Large-Scale Benchmark of the Kurdish Handwritten Lines Dataset for Low-Resource Central Kurdish (Sorani)},
  author={Ibrahim, Mohammed S. and Ahmed, Hevi and Hussein, Yassin and Rashid, Tarik A. and Fattah, Polla},
  year={2026},
  publisher={Mendeley Data},
  version={2},
  doi={10.17632/ffjb8kfb7m.2},
  url={http://doi.org/10.17632/ffjb8kfb7m.2}
}

Licensing

This dataset is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0).

Contributions

Contributed by the Artificial Intelligence and Innovation Center (AIIC), University of Kurdistan Hewlêr, and Salahaddin University-Erbil.