May 30, 2019 by Greg Meckbach
Do your workers spend hours typing information from paper forms into the computer?
There are ways of automating this.
“Nowadays, you probably have documents coming in and you put them on the scanner. When you are going to put them into whatever system, you want to be looking for the system to have some ability to do optical character recognition,” said Christopher Wynder, director of product marketing for capture and original equipment manufacturer products at Open Text Corporation, in a recent interview. “That’s kind of tablestakes for many of the word processing we see nowadays.”
OCR, or optical character recognition, is when the computer can read a paper document and enter the text into software – such as a word processor.
“Wherever you are housing those documents should have the ability to do the extraction of the typed and electronic letters,” said Wynder.
Before he joined Waterloo, Ont.-based Open Text, a multinational business software vendor, Wynder was a technology consultant. One of his clients was a small brokerage that received many paper documents.
“They had access to a system to deposit the documents but it didn’t do OCR. So what they ended up doing is hand-typing everything,” Wynder said of that brokerage. One of the solutions he worked on, for that brokerage, was a method to get the information from a paper document into a business software system without a lot of human effort.
“That’s exactly what a lot of [Open Text] customers are craving – the idea that we should take a paper document, scan it in, do the OCR and do the document classification.” The goal is for the underwriter to do their job and not have to look at the document in the capture system.
Insurance professionals may need to manually read and re-key in information when the OCR has low confidence that it correctly recognizes letters.
“The one thing we know about handwriting is, there is a lot of variability,” said Wynder.
“If it’s a lower case H, if there is just a little bit of pixelation, that H might look like an N to a machine.”
With OCR, the software compares what is on a paper to a template showing letters – with the aim of figuring out what the character is.
But sometimes the computer will go back to the user and say, “not sure. It could be an H. It could be an L. It could be an M,” said Wynder.
This is where machine learning and AI can come into play.
“No one wants to spend their day figuring out if that was an H or just somebody’s lazy N,” said Wynder.
ICR, or intelligent character recognition, is when the software looks at handwriting and looks at the surrounding pixels if it does not recognize a character.
The machine learning could look at an entire word and figure out whether it correctly guessed one particular letter, Wynder suggested.
The holy grail is to relieve the worker of having to look at handwritten letters and figure out what they are.
“The goal is to focus be on the high- value things which are defining whether that document provides the necessary information and defining whether there is enough information to do the approvals. These kinds of things that are still very much a human task,” said Wynder.