Gregory's Server@PJ_Evans @internetarchive I also want to point out that scanned books are a much poorer reading experience than e-books - it's just a photograph of book pages, with maybe some un-reviewed OCR text behind it. Fine perhaps for some searching, discovery of a proper noun in a large amount of text, but we can't promote careful reading with this technology. I hope that someone is working on a way to turn these images into actual book text.
|
3 comments
@PJ_Evans @internetarchive The text can be really garbled. Note that it is offered to the visually impaired and I would like to hear how well it is working for them. When I look at it, it's a mess for things like tables of contents; plain text renders better, but there are still errors. Could something be set up where the text could be corrected by humans?  |
@kcoyle @internetarchive
That's why I prefer the scanned image to OCR'd test. (I had a job, for a while, fixing OCR'd text in insurance laws. Some of the output was actually funny, like "legal obligation" becoming "lethal obligation" and "District" getting scrambled in to "Omelet".)