Package contains 450 page images and ALTO XML files for each page, with the proofreading done by the Finnish native speakers.


Description

The pages of fraktur range from the year 1836 until 1910. The package can help in creating own postcorrection algorithms for OCR text recognition.

There is also an Excel file for all of the 471 903 words, which contains result given to the word by Tesseract and FineReader. If a tool hasn't found corresponding word, then the given cell is empty, so select the words in the Excel, which you need.

NB! The ground truth package does not contain the data for the 1918 due to copyright reasons.

User interface
Data downloadshttp://digi.kansalliskirjasto.fi/opendata  
API-
LicenseTerms of use (in Finnish).






Content typePageimages, ALTO XML, metadata
LanguageFinnish
Data statusPrimary source
Size450 pages
Update frequency-
Relationships-
External informationArticle: Creating and using ground truth OCR sample data for Finnish historical newspapers and journals (PDF file)
Acknowledgements

Digitalia project (http://blogs.helsinki.fi/digitalia/)


Leverage from the EU 2014-2020European Union
Contact information[email protected]
Star rating-