Translating Documents From Reels and Microfiche/film

[Stubble]

Fellow forum users,

Hi, it's me, Stubble. I wanted to know what solutions you have used generating translations of microfilm/fische from German to English. Currently, my plan is thus;

1) Set up a box that will use OCR to convert the reels to text.

(OpenCV for example)
2) Translate the documents to English.

(MarianMT for example)
3) Winnow the output.

(Can be automated or manual)

Anybody done this before? Am I barking up the wrong tree? Is there an easier way to do this that anyone can think of. I'm currently looking at a very large volume of files to winnow, sort, collate and examine.

Thank you for your time,

Stubble

[Archie]

If you know a bit of German, you might be able to skim through the German pages and look for keywords like Juden. And then only transcribe and translate select pages.

See here for some tips on learning German
https://codohforum.com/viewtopic.php?t=131

A few years ago, traditional OCR was far from perfect, especially for old documents with not so great image quality. But this technology should have improved a lot with LLMs since they are so good at probabilistic word prediction. I don't know if there are any good tools that will do LLM correction on a large number of pages for free.

If it becomes too much, you might try to narrow things down a bit. Like if you know what dates are most likely to be important for the question you are looking into, you can jump to those dates. Some microfilm collections will have a finding aid on the first roll. That can help you figure out what material is most likely to be relevant.

[Stubble]

Thank you Archie, I appreciate it. I'll look at some of the specific dates and put together some documents to translate and see if I can get some gold out of them.

I also sent you a PM. Check your inbox for a more expansive explanation.

Time to put the flippers and snorkel back on. I'll be around.