261–267 of 267 posts
Pretty easy to get started after signing up at login.gov
then https://www.archives.gov/citizen-archivist/get-started-trans... with video: https://www.youtube.com/watch?v=dwQ5pEWWFY8
The support forum is https://historyhub.history.gov/citizen_archivists/ to ask questions
Earlier quoted context omitted.
What? I pulled one page out of the image set and tried to get GPT 4o to transcribe it. I wasn't just using the easy example from the original article, it's an easy example to draw people into the idea of participating in the volunteer effort. If it were one of the inscrutable documents people would be more likely to be put off the effort. Did the link in my comment not take you to a single page (I just tested it in i…
No, if I follow the link in your comment, I get a very different image, this one: https://s3.amazonaws.com/NARAprodstorage/opastorage/live/17/... (page 8 of "Revolutionary War Pension and Bounty Land Warrant Application File W. 7785, John Hopper, N.C.") I agree that your description of the image the link shows you, which appears to be page 52 of the same file, makes sense. I can read ... some of the handwritten words…
Anyways yes that handwritten text is an example where LLMs just cant hack it and people seem to be able to. There's a pretty thorough transcript of the upper handwritten portion of the page I was referencing available from a user. It's a great example of why you can't just throw an LLM at problems like this. At best they're a tool people can use to transcribe loads of them quickly but it still needs to be hand checked for accuracy, completeness, and relevance.
Earlier quoted context omitted.
I don't know about this project, but I can easily find thousands of images that gpt-4o can't read, but a human expert can. It can do typed text excellently, antika-style cursive if it's very neat, and kurrent-style cursive never.
For straightforward reasons, I am commenting on this project, not the space of all possible projects. I did try, once, to get 4o to decode the Zodiac Killer's message. It didn't work.
Earlier quoted context omitted.
Please don't link to wikipedia definitions of elementary terms. It's condescending and obnoxious.
> Please don't link to wikipedia definitions of elementary terms. It's condescending and obnoxious. I respect what you have chosen to contribute to this conversation, but neither need nor seek your approval of mine.
Earlier quoted context omitted.
For straightforward reasons, I am commenting on this project, not the space of all possible projects. I did try, once, to get 4o to decode the Zodiac Killer's message. It didn't work.
The point is that gpt-4o can't read cursive very well. The chain of thought reasoning doesn't seem to help at all. It's a thing the model doesn't generalize to, that's relevant to this discussion.
I think people may be kidding themselves a little bit about this stuff.
To tptacek and other guys who seem to have unwavering trust in OCRs/LLMs, as well as to opposite party who think that technology is not there yet — you are all partially right, but somehow fail to hear each other while also spending time on baseless arguing instead of factual examples and attempts to find common truth. Can it be used to greatly simplify efforts by getting through boilerplate? — Yes. Should the result…
Earlier quoted context omitted.
The point is that gpt-4o can't read cursive very well. The chain of thought reasoning doesn't seem to help at all. It's a thing the model doesn't generalize to, that's relevant to this discussion.
I'm sure you can find something it does a bad job on (have you, though?) but it's done swimmingly on some really horrendous 18th century examples I've found. Among other obvious things, it translates the Revolutionary War cursive that leads off this article. I think people may be kidding themselves a little bit about this stuff.
But every time I find something that I think "This one is surely easy enough!" it's usually wrong. US revolutionary war cursive is on the easy end; it's quite similar to modern cursive, and what is of training data in the datasets is probably a lot like it, if not actually in it - historians who test this complain about overfitting. Which I believe, because as I said, I see that it's bad at generalizing.