Hello colleagues,
I have the following problem: I train tesseract on the particular font, font-style, color, character casing, etc. I use for training real samples from the documents I want to scan and OCR. Thus I reduce the ambiguity coming from font differences. This works pretty well but with one exception. Unfortunately few letter combinations sometimes could get printed without enough spacing and become agglutinated. For example I have the letter "T" followed by "Y". When these two letters are slightly agglutinated one to another tesseract fails to recognize them correctly. How to solve this? Do you think that providing DangAmbigs file during the training will help? Note that my case is not typical ambiguity of two letters looking like a single another letter. How do I know all possible letter pair that potentially could become ambiguous when agglutinated? I have the following idea: add few fictive letters in the training alphabet. For example I could add "TY" as a single letter and use some not occupied Unicode character for this fictive letter, e.g. $. Later when tessearct finds $ as result of OCR operation I could replace back $ with the letter sequence "TY". This should work but I still believe there should be more simple way to overcome such recognition errors. Does anybody have experience with this problem? Svetlin Nakov Development Manager Intelligent Software Consulting (ISC) --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/tesseract-ocr?hl=en -~----------~----~----~----~------~----~------~--~---
