Dear fellows, I have somewhat a difficult problem for me. I googled many times and found many hints and tipps but not a really working methode for me
I'm making it short as possible: I want to OCR an old german dictionary for my master-thesis in linguistics. The available traineddata of German (Fraktur) on github is very good, but one type/ style of German Fraktur is missing and won't be recognized properly. The following picture shows the "problem": the first word is "Imporös" followed by a definition where you can see the word "Imporösein". You can see the different types of the fraktur if you look at the two words, but the difference is most visible if you look at the two I's at the beginning of these words: <https://lh3.googleusercontent.com/-wM1_X11Fy6g/V0Rd9YPzenI/AAAAAAAAAac/NpAdpWGRadEjASckuljk9-EjZq7LW0wUACLcB/s1600/example_fraktur.jpg> Now for every letter of the alphabet there are at least two different styles of letters. The normal traineddata of deu_frak is able to only recognize ONE of them. Now I want to train Tesseract to be able to read both of them. First I thought I should create "a new language". So I started with Aletheia and then proceeded in Franken+. But the traineddata of deu_frak on github is not bad, I just need to add some glyphs/letter. Otherwise I need to start a complete new langdata but that's going to be too much work since the dictionary is very complicated and needs a lot of manual correction in Aletheia. I have downloaded the langdata on github that are needed (the are in the folder "frk") but I don't know what to do with them. How can I add another letters/glyphs to be recognized correctly? I was also confused when I unpacked the original traineddata "deu_frak" with Tesseract of Aletheia that I get somehow complete different files. If needed, I can attach the folder containing those files. I think it's not really helping to solve my problem that I'm working on Windows? Well I'm actually just a linguist so I have worked my way through all of that by myself but somehow I just need to be able to re-train that already existing and good traineddata of "deu_frak". Maybe someone here could help my out, that would be just too great! Thanks a lot! Regards, Sebastian -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/2f5e36ad-9a22-42a1-8094-c09b06d01522%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
