PDA

View Full Version : Document Mangement/OCR advice

WillBradTMA
03-23-2009, 07:22 AM
Can anyone give me advice on a solution that my company is looking into (never had to deal with it beyond a passing knowledge level).

They are looking for a solution to be able to scan a document to a PDF file as well as capture data off of that document to push to a SQL database.

So far the only thing I have found that is a solution from Brainware (http://www.brainware.com/data_capture.php). However it seems like it will be more than our current budget would allow for.

Here is the overall process

1. Document is scanned and named according to the UID on the document (in this case model and serial number)
2. Specific fields (possibly could be zonal, the issue would be different formatting of document for each model, we currently have 21 different models) need to be "read" and the handwritten text and numbers pushed to a SQL database


TIA
mjp
03-23-2009, 08:07 AM
That Brainware app is interesting. I'm not surprised it's expensive if they've actually come up with a way to do relatively error-free OCR. Especially with hand written data.

I don't know of anything else that attempts to do that, but maybe someone else will.
WillBradTMA
03-23-2009, 01:57 PM
Well if you interested in what they say I can let you know what they say.
mjp
03-23-2009, 02:37 PM
Sure, let us know.
WillBradTMA
03-24-2009, 08:40 AM
Sure, let us know.

Okay I assume you like to know cost and how good are the success rates at translating handwriting into text. According to one of there sales guys its better than 95% under prefect conditions. For the conditions we have (manufacturing ie dirty, greasy, sometimes wet documents) they are quoting around 60% (note they have no seen a actual sample yet).
mjp
03-24-2009, 10:20 AM
Yeah, that's about what I would expect for any hand written OCR, 60% accuracy at best. There are too many variations in handwriting styles to do any kind of effective OCR.

It's an interesting task to try to pull off, but I think you may end up scanning the documents to PFD and doing human data entry from that. If you have to check automated OCR for errors, a lot of times it turns out to be quicker to simply do the data entry manually.
WillBradTMA
03-24-2009, 01:13 PM
That is the plan is to scan them to PDF and have the OCR write the data.