Just to be clear, I am extremely new to ABBYY - I just finished basic training, and started playing around with the first of our actual forms this morning. I've been banging my head all morning trying to figure out what's wrong: I loaded a form into the document definition editor, and let it autodetect. I then configured all the fields, which took a while, so I'd really like to not have to do that again on the same form. Then I tested it against some filled-in samples, and determined the autodetected separators wouldn't all work perfectly, due to variation in scanning, so I was going to play around with fewer, more specific static elements. However, when I delete some of them, it tells me the definition is no longer valid, with the following errors (some of them I can delete without error):
"Some elements used for Document Definition matching cannot be matched"
"Fewer anchors in image than in Document Definition"
"More black separators in image than in Document Definition"
If I put the anchor I deleted back, the errors go away. I have no idea why this is happening, or whether it's expected? It certainly doesn't seem expected that removing one anchor would cause errors like that, especially if the anchor removed wasn't even a separator?
More information: I deleted every anchor on the page, and the error went away. I added one text anchor, no errors. I added one traditional black box anchor, no errors. I added the second traditional black box anchor, the errors came back. I think there's something wrong with this doc def? (Hopefully something there's a way to fix, as I spent a pretty good chunk of time setting up all the fields...)
Ironically, the only other hit I could find for that error is this thread, which isn't quite the same: http://www.capturedocs.com/thread/document-definition-classification-problems - funny enough, that is also what I'm trying to solve, so as an addendum, I also wanted to confirm that it's true, you can't have a single fixed form for multiple languages even if the layout is identical, just the words on the page differ? Or was that answer assuming that the layout may differ slightly between languages as well? The documentation seemed to indicate that if you added a text anchor and said to use it for identification, but didn't specify a value, that it would just use the existence of text at the location rather than the exact text found there, but it wasn't completely clear?