ANALYSIS AND COMPARISON OF OPEN-SOURCE TOOLS FOR OPTICAL CHARACTER RECOGNITION IN BUSINESS DOCUMENT PROCESSING
Annual of Konstantin Preslavsky University of Shumen, Faculty of mathematics and informatics · 2025
DOI: 10.46687/hive1644 ↗Kurz gefasst
Fachartikel in der Zeitschrift Annual of Konstantin Preslavsky University of Shumen, Faculty of mathematics and informatics, 2025. Verfasst von STOYAN KONSTANTINOV.
Abstract (Originalsprache)
This paper compares leading open-source tools for optical character recognition (OCR) in business document processing. Five platforms—Tesseract, EasyOCR, PaddleOCR, docTR, and Donut—are evaluated for architecture, functionality, accuracy, and resource efficiency, with focus on Bulgarian-language support and intelligent document processing (IDP) capabilities such as layout, table, and key-value extraction. Findings show PaddleOCR offers the best balance of performance, usability, and cost, while hybrid OCR–language model approaches improve flexibility, and Donut fine-tuning achieves the highest accuracy for specific document types. Open-source OCR and IDP tools enable automation of document workflows while preserving data sovereignty and avoiding vendor lock-in.
Zitieren
APA
KONSTANTINOV, S. (2025). ANALYSIS AND COMPARISON OF OPEN-SOURCE TOOLS FOR OPTICAL CHARACTER RECOGNITION IN BUSINESS DOCUMENT PROCESSING. Annual of Konstantin Preslavsky University of Shumen, Faculty of mathematics and informatics. https://doi.org/10.46687/hive1644
Vancouver
KONSTANTINOV S. ANALYSIS AND COMPARISON OF OPEN-SOURCE TOOLS FOR OPTICAL CHARACTER RECOGNITION IN BUSINESS DOCUMENT PROCESSING. Annual of Konstantin Preslavsky University of Shumen, Faculty of mathematics and informatics. 2025. doi:10.46687/hive1644
BibTeX
@article{KONSTANTINOV2025,
author = {KONSTANTINOV, STOYAN},
title = {ANALYSIS AND COMPARISON OF OPEN-SOURCE TOOLS FOR OPTICAL CHARACTER RECOGNITION IN BUSINESS DOCUMENT PROCESSING},
journal = {Annual of Konstantin Preslavsky University of Shumen, Faculty of mathematics and informatics},
year = {2025},
doi = {10.46687/hive1644}
}
EUVIMED