1

I have two pdf files:

  • original image pdf where the pdf file is just an image of the text

  • ocr pdf produced by OCR, using ABBYY FineReader 11

these are the process i did

  1. i have a image pdf then i use abby to perform OCR function

  2. then i convert my pdf to scanble pdf in abby transformer 3.0

  3. then i compare both the source file and ocr file using acrobat but all those hyphens are in error how to skip those hyphens while searching ?

  • Your question is not too clear. Do you get the second file by applying ABBYY FineReader 11 to the first file. Is the second file just standard PDF text (no image). Why do you want to compare? What do you call an hyphen error? Is an erreur just a difference between the files? Actually, I am surprised Acrobat will compare an image and a text, but I am no expert on Acrobat. BTW, there are PDF files that contain both text and image (in a correlated structure), so that one can search the text, while retaining the original image for accuracy. Are you using one of those? – babou Feb 4 '16 at 15:26
  • i update the question please check those – User_3535 Feb 9 '16 at 8:47
  • You should probably add the tag hyphenation in your question. Also scanble does not seem to be an existing word. I guess you mean "scannable", and there is a tool with that name. Is that what you actually mean? – babou Feb 9 '16 at 11:28
1

If I understand your question correctly, you are asking for a way to automatically double-check an OCR conversion from image to text. I'm afraid that is not possible: checking would require conversion from image to text via OCR, which I'm sure you can see is circular. If you want to proof an OCR conversion, you'll have to do it the old-fashioned way. At least it's faster than typing the whole thing out (and then proofing it again)!

  • hi .,, these are the process i did 1. i have a image pdf then i use abby to perform OCR function 2. then i convert my pdf to scanble pdf in abby transformer 3.0 3. then i compare both the source file and ocr file but all those hyphens are in error how to skip those hyphens while searching ? – User_3535 Feb 9 '16 at 8:43
0

I cannot answer you with certainty because I do not have the tools yo use. Hence I am not sure what kind of file they create (PDF can come in different flavors). Furthermore, besides the basic character recognition, OCR may or may not involve various steps, often based on linguistic analysis, intended to improve the result. Hyphen removal is one such step, since there is usually no point in keeping hyphenation in the resulting OCRed text.

See for example this question: How to remove hard hyphens? or this Master thesis on Dehyphenation.

Comparing images is sometimes possible, but unlikely to give you easily usable results for textual purposes.

Thus one must assume that you used one tool (ABBYY FineReader 11) to get a first OCRed document, and that your comparison did a second OCR on the image to compare it with the first result.

This may make sense, and help identify some possibly erroneous locations in the OCRed text (though I do not know that it is used, as there may be other ways). It is also likely that both OCR process will sometimes agree on making the same mistake, which then remains undetected.

Now it is possible that one OCR system does hyphenation removal while the other does not. Then the comparison of resulting files would show differences wherever there was an hyphen in the original image. BTW, does acrobat find (mostly) extra hyphens or missing hyphens?

To do away with the problem would require that both OCR steps do the same regarding hyphenation, either keep it or remove it. I would expect that some OCR algorithms can be configured to do either which might solve your problem.

Note however that hyphen removal may sometimes be ambiguous, requiring somewhat arbitrary choice. Thus if both OCR systems used for double checking do hyphen removal, they may sometimes disagree on which hyphens should be removed, and there is not much one can do to avoid this.

Another way to avoid hyphens is to write a piece of software that will filter out reported hyphen differences. How to do that depends on the context you work in.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.