Typically, depending on the quality and typeface of the original, the accuracy with which it was laid on the scanner platen (twist or skew of the page), shading due to the gutter of the binding (if a book), one might get an accuracy of 98%. However, one point to make: no matter what OCR application one uses there are errors. Using linux there are quite powerful OCR tools which do not suffer from the above problem (if set correctly). To get the text out of the little frames into an unformatted stream can be quite a challenge.
This is because they often default to attempting to recreate the exact layout of the original and may be putting text in little frames, rather than outputting it as an unformatted stream.
The output of most OCR programs is, as esperantisto says, "a bit weird".
0 kommentar(er)
