Making Old Kurdish Publications Processable by Augmenting Available Optical Character Recognition Engines
Authors
arXiv
April 9, 2024
Abstract
This 2024 arXiv preprint adapts Tesseract 5 for historical printed Sorani Kurdish publications using transcribed lines from pre-1950 material. The abstract reports 1,233 line images with transcriptions, while the currently inspected partial release contains 1,203 TIFF images and 1,203 text files. The paper reports 0.755% CER in its Tesseract workflow and 84.02% average character accuracy in a separate Ocreval assessment.
Keywords
OCR
Sorani Kurdish
Historical Print
Tesseract
Preprint
Related Datasets
KurdishOldTextsOCR
2024
1,203 TIFF images and 1,203 text files in the inspected release
TIFF, TXT
Ground truth for OCR of historical printed Sorani Kurdish publications. Direct archive inspection found 1,203 TIFF images and 1,203 text files. The associated paper reports 1,233 line images, so the …
Citation
Yaseen, B., & Hassani, H. (2024). Making Old Kurdish Publications Processable by Augmenting Available Optical Character Recognition Engines. arXiv:2404.06101.