"Based on an input of human transcriptions – a few dozen pages will suffice – the computer develops a reading model that can be more precise, and that will cert

"Based on an input of human transcriptions – a few dozen pages will suffice – the computer develops a reading model that can be more precise, and that will certainly be quicker, than handing the transcription work out to humans. " I really like the idea of per-document OCR models. So much of Europe's culture and history is locked in latin books that have been scanned onto the archive but not successfully OCRed. Progress on this front would be nice! (As someone ignorant of both European and Chinese paleography, China's historical text digitization projects seem to be far in advance of Europe's - I was able download 200k classical chinese poems like it's nothing in a nicely formatted json archive on GitHub, for instance).