Hybrid OCR/HTR Framework for Quality-Aware Text Extraction from Heterogeneous Digital Archives: Source Code and Experimental Results
Annotatsiya
This software package contains the source code and experimental results associated with the study “A Hybrid OCR/HTR Framework for Quality-Aware Text Extraction from Heterogeneous Digital Archives”. The implementation provides a quality-aware and document-type-adaptive OCR/HTR pipeline integrating document quality validation, type-specific routing, image preprocessing, PDFMiner-based text extraction, Tesseract 5 OCR, domain-adapted TrOCR HTR, text normalization, and evaluation modules. The archive also includes experimental outputs, baseline and ablation results, statistical evaluation files, processing-time results, and generated figures used to support reproducibility of the study.
Hali tarjima qilinmagan