Introduction to Transkribus and the new User Interface - Webinar (English)

Discover the capabilities of Transkribus in this comprehensive webinar. Learn how this AI-powered tool simplifies the transcription, annotation, and management of historical documents. Explore its unique features, including portable scanning solutions, custom AI model training, and new digital publishing options with Transkribus Sites. Perfect for historians, researchers, and archivists, this introduction to Transkribus is a must-watch for anyone working with historical texts. Dive into the future of unlocking the past!

[cosmiccrunch]

A feature I would absolutely love to see is the ability to recognize and ignore page headers and page numbers during the recognition process. An example use case is publishing of important old public domain works in different languages.

So let's say I want to run the exported transcription through a service like DeepL to translate it to English. DeepL uses context to determine the most accurate translation. If there is a page header (chapter title, book title, page number) between the body of text on one page and the next it interferes with the translation.

The only currently remedy is to manually split the region and delete the header region for every single page of the document. Productivity would soar with this one feature alone in my use case, which equates to more more pages processed through your service. A win-win situation by far.

Beyond the above, the only other feature I'm missing for this to be the cats meow is automatic recognition of footnotes and marginalia, with the option to export these regions separately from the main text.

[hillarymcnerney]

In the training section, she showed a German newspaper in Fraktur font, which is exactly the model I need, however this does not show up anywhere when I search for models. How can I find and use this model?