NCR Corp. is upgrading a processing system that lets financial institutions replace paper documents with computer images.

The enhancements, which will be announced today, are designed to let the system recognize a wider range of characters in documents stored as digitized images.

Optical character recognition, or OCR, is most useful to financial institutions for processing loan applications, where manual keying of printed information is perhaps the most time-consuming step.

"Banks and other companies have a great need for a single OCR system that will recognize many different types of characters, and it appears now, for the first time, that they will have that," said Mary Bamford, an industry analyst at BIS Strategic Decisions, an image technology consulting firm based in Norwell, Mass.

Most Systems Fall Short

Optical character recognition is considered an essential component of any document imaging system marketed on the basis of improved productivity. But few companies sell a software package that can read the wide range of handwritten and machine-printed characters on checks and other bank documents.

As a result, few document imaging systems have achieved the claimed potential of reducing document handling and processing costs by 30% to 40%.

Many experts believe that is about to change. The new software from NCR, which is a subsidiary of American Telephone and Telegraph Co., and a similar package released by REI Software earlier this year have begun to expand the range of characters that can be read by a single software program.

"To get this kind of capability a year ago, a bank would have had to string three or four packages together," said John Lynch, director of network and strategic marketing in NCR's U.S. group.

"Each software package would have to be invoked individually by the user until the characters could be recognized."

By contrast, the new NCR software automatically performs three kinds of algorithm tests, permitting recognition of machine-printed characters, magnetic ink lines on checks, and bar codes, as well as handwritten information.

Less Work for People

A computer operator is needed to key information into the system only if all three tests fail to identify a character.

First National Bank of Maryland, Baltimore, is the first financial institution to purchase the system.

But bank officials said the system has not been fully installed yet; they declined to comment on the expected productivity gains.

The new software, which NCR will unveil in Anaheim, Calif., at the annual conference of the Association for Information and Image Management, will be generally available in about 60 days, officials said.

A complete system, including the character recognition software and Unix-based computers, costs about $800,000 for a 20-workstation configuration.

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.