Blogs

If only we could be more predictable.

I include myself in that statement, of course.  It's very hard for us, as people, to do the same task over, and over, and over again, and to do it consistently.  Maybe it is our curious nature, that our minds can't help but wander.

This becomes obvious when you ask a person to key-enter a stack of papers.  There will be errors, and their frequency will likely increase over time.

Obviously, we hate errors.  Errors in data are unacceptable.  If you can't trust your data, what good is it?

What do we do to fix this?  Automate!

There are a few common methods for automating Metadata capture.  I'm listing them in order of pereference based on data correctness.

• Database lookup.  Is the metadata already in another database?  Look it up and link it.  If your personnel data is already stored elsewhere, use a lookup at scan-time on the employee's unique ID number to pull the rest of the necessary data into the capture record.  One other thought for this process - be sure you handle propogating changes in the source database over to your imaged metadata.
• OCR.  This only works for certain document types (like forms), but in those cases it can work very well.  You will also need clean source documents - faxes may not work.  Using Zone-based OCR, document recognition, and forms validation features can nearly completely automate the metadata capture process.  The scan operator MUST validate the capture, though, as OCR isn't perfect.
• Field validation.  Sometimes you can't get away from a person entering the data.  Wherever possible, constriain that input to make it as consistent as possible.  Use field characters and regular expressions to limit the data entry to the acceptable formats.  Limit character case wherever necessary.  Add as much logic as possible here to help cull out the errors.

None of us is perfect, and none of us can provide perfect data, although that is our goal.  Use as much technology as possible to help aproach that goal.