Reconciling the "Gebrauch" (use) column of the data with Wikidata worked well when the values followed the usual pattern.
It was more difficult when the "use" didn't make sense without the object title to guide us.

112
Experience
Dribs
3 days ago
The VMZ's catalogue database can export the images of a subset of items (the result of a query, like "All items with question marks in their title"). Unfortunately, it can only export the images as a PDF. I wrote a quick script to extract all the images from the PDF and save them with the item id as the file name: https://gist.github.com/bellisk/c0d80fda6d21d14d45a37fffa4f1a333
4 days ago
We have set up an instance of eScriptorium that's available online, and we've found a collection of images that might be relevant to our question. Now we're training an OCR model on those images in eScriptorium.
4 years ago