Re:Inside the OED: can the world’s biggest dictionary surviv...
by ComputerGuru on Sunday February 25, 2018 @11:30PM
(#16463138)
OCR has come a _long_ way since 2005. I was an early adopter around 2000, and the results scared me away from OCR for thE better part of two decades. I recently revisted it with no special approach, just trying Adobe Acrobat Pro’s OCR module and was very pleasantly surprised with the progress that has been made.
Adobe (and I’m sure many others) now preserve the image but place text behind it in a hidden layer or else use a pseudo ML process to create a font from the scan and fit it to the text with high accuracy (and low binary sizes).
Might be worth revisiting.
Edit: just saw the full mailing list url (on phone, hard to see anything) and realized who I am replying to. Hi! I have missed your writings! I recommend “only a constant factor worse than optimal” all the time to people.
Adobe (and I’m sure many others) now preserve the image but place text behind it in a hidden layer or else use a pseudo ML process to create a font from the scan and fit it to the text with high accuracy (and low binary sizes).
Might be worth revisiting.
Edit: just saw the full mailing list url (on phone, hard to see anything) and realized who I am replying to. Hi! I have missed your writings! I recommend “only a constant factor worse than optimal” all the time to people.