[Balrog49]

I have some printed system notes that I want to convert to an online format. Does anyone know of an OCR tool that can recognize suit symbols and that does a decent job with hand diagrams?

[axman]

 Balrog49, on 2014-March-18, 14:30, said:

I have some printed system notes that I want to convert to an online format. Does anyone know of an OCR tool that can recognize suit symbols and that does a decent job with hand diagrams?

I've used omnipage [14] for several years at 300dpi scan. it has a train feature and I train cdhs as {}[]. There are fonts that recognize those characters as bridge symbols. however, what is do is to then search and replace for the character map symbols. Sometimes the ocr skips the symbols and you have to edit them back in [usually best done during the proofing process.

I also find it desirable to recognize bridge deals as spreadsheet tables rather than regular text.

good luck

[Balrog49]

Thanks for the response! I've been trying to use your method with Omnipage 18 but I can't get training to work. If I scan a page at 300 dpi and draw a text zone around a single suit symbol in the page view, "train character" is greyed out.

It seems to me that Omnipage should be able to recognize any character in any specific font, including Arial, which has suit symbols. But according to their support line, it can only recognize "standard" characters. What a disappointment!

Am I missing something?

[axman]

 Balrog49, on 2014-April-07, 15:06, said:

Thanks for the response! I've been trying to use your method with Omnipage 18 but I can't get training to work. If I scan a page at 300 dpi and draw a text zone around a single suit symbol in the page view, "train character" is greyed out.

It seems to me that Omnipage should be able to recognize any character in any specific font, including Arial, which has suit symbols. But according to their support line, it can only recognize "standard" characters. What a disappointment!

Am I missing something?

training is done after ocr on the doc. you highlight the 'text' that is incorrect and right click to get a menu, select train to get a dialog. that is where you use {}[] as a euphemism. then, in your word processor use find/replace [to see the suit symbol]. in other words omnipage doesn't train to get suit symbol- it finds suit graphics and ocrs a character you have chosen. you will want to save the training file for when you ocr similar quality documents [I find that the program tends to get confused between different type faces so you will want to train to different training files. you can if you wish try old training files in hopes of not needing to train for new projects].

[Balrog49]

Success! Thanks!

I didn't think the suit symbol character combinations in the text editor were consistent enough to train but apparently they are, although some number-symbol combinations produce several different results. I can now save a Word file containing 1C, 2D, 3H, 4S, etc. and I have a macro that transforms those into red and black suit symbols.

And I didn't realize that you have to manually save the training file each time you repeat the recognition step.

It's still going to be a lot of work to get everything formatted correctly but at least the job can be done. I was ready to blow it off.