[quote user="harryLondon"] One thing to be very careful about when using free ocr . often, the character recognition is relatively poor, but is enhanced by comparing the results with common dictionary words to decide what a fuzzy character within the word is most likely to be. This works well for paragraph text but is likely to give errors in numeric data that appears withing the text because a 6 may have only a few pixels difference from an 8 and there is no equivalent way to decide if one is more likely than the other. People are often fooled seeing the apparent accuracy of the recognised text into assuming that the result is a faithful reproduction of the document, but in practice text from OCR needs to be proofread even more carefully than typed-in text. [/quote] I agree, the accurate of ocr can't be 100%, all the ocr software may have this problem.
↧