Austrian Newspapers is a ground truth data set created with Transkribus from Austrian newspapers by the Library Labs of the Austrian National Library (Österreichische Nationalbibliothek). See this publication for details:
Günter Mühlberger, & Günter Hackl. (2019). NewsEye / READ OCR training dataset from Austrian Newspapers (19th C.) [Data set]. Zenodo. http://doi.org/10.5281/zenodo.3387369
The original data set was published under the Creative Commons Attribution 4.0 International license.
This repository started with the original data set and now tries to enhance it in several aspects to make it more suitable for model training of OCR software.
The original transcription uses level 1 according to the OCR-D transcription guidelines (German).
This is sufficient for many applications, but there is also a high need to have more detailed transcriptions. Therefore the transcriptions were enhanced with the long "s" and special unicodes for fractions, and similar enhancements might be added in the future. It is still possible to get level 1 transcriptions by simple search and replace operations.
The initial enhancements were added by UB Mannheim who also fixed some transcription errors in the data. See https://github.com/tesseract-ocr/tesstrain/wiki/AustrianNewspapers for a report about the model training at UB Mannheim.
New contributions (issues, pull requests) and reports about successful trainings are welcome.
Unpacked original data (PAGE XML, TIFF) from ÖNB with some transcription texts enhanced and fixed
Extracted line pairs (image and ground truth text)
Note: Many updates in gt still have to be transferred to the PAGE XML files.