Gregory's ServerIf machines are so much better than humans at captchas now, can we have the book OCR failures ones back? Please? Because there's a lot of archives out there and uh, we still can't OCR Akkadian; we can't even reliably OCR printed materials from before about 1980. Let alone handwritten. Recaptcha was a force for good before it got, ironically, captured.
 1 comment
|
@feonixrift There’s a period from about 1960 forward where all us dot gov docs were in that awful all-caps format that we humans find unreadable. however, the fed had OCR text that could read all-caps text really well. those papers are out in the world now. I actually have a book here some place from 1959 about how they accomplished it. If US Census was involved, that tech is out there someplace open source, because that’s what they do. So we might have something to start from. Might be in cobol or something though :)
@feonixrift There’s a period from about 1960 forward where all us dot gov docs were in that awful all-caps format that we humans find unreadable. however, the fed had OCR text that could read all-caps text really well. those papers are out in the world now. I actually have a book here some place from 1959 about how they accomplished it. If US Census was involved, that tech is out there someplace open source, because that’s what they do. So we might have something to start from. Might be in cobol...