Optical character recognition used to “read” documents still has challenges when it comes to identifying data on forms, but these can be addressed, according to Jonathan Kunkle, president of LenderLive Network’s
Most OCR technology reads documents in standard ways, identifying certain types of data based on where it is located on those docs.
But some docs such as bank statements, which are collected and analyzed as part of a loan origination or modification application process, are not designed in standard ways such that the same information is in the same place on all of them.
“Most OCR is zonal-based,” meaning the technology will expect to find a particular type of data in a particular place on the document, he said.
No single OCR tool can solve this problem, Kunkle said, adding, “so what we’ve done is employed a number of tools” so that the automation is “reading the document more like a human being.”
This OCR is more “intuitive to the form” and uses a more “relational” approach, identifying what type of data it is based on where it is relative to another type of data on the document.
While not all forms are the same, companies do tend to put certain types of data near certain other types of data, he said.
