BALL'S TARGETS SPEECH RECOGNITION, IMAGING

KANSAS CITY, Kan. -- Development of a cashier training tool using speech recognition and imaging technology, to get under way at Balls Food Stores here in the coming weeks, could lead to design of a voice-activated point-of-sale system.The retailer and its vendor partner are building a data base that will contain computerized images of some 400 produce items, said Brad Riffel, director of information

KANSAS CITY, Kan. -- Development of a cashier training tool using speech recognition and imaging technology, to get under way at Balls Food Stores here in the coming weeks, could lead to design of a voice-activated point-of-sale system.

The retailer and its vendor partner are building a data base that will contain computerized images of some 400 produce items, said Brad Riffel, director of information systems operations. The data base will support a cashier training system focused exclusively on produce identification and pricing accuracy.

"We want to increase customer confidence in our pricing and our ability to identify all these different produce items," he said.

The retailer's own research uncovered an 11% error rate in produce identification. "So we know there's a problem, particularly with the more exotic items that are purchased less frequently," Riffel said.

"The industry as a whole knows this is a problem," he added.

Design and development of a beta version will continue until the fall when Balls hopes to launch an in-store test of the cashier training tool.

"That's as far as we've contracted to go. Down the road, we hope we can interface this technology to the front end so cashiers can ring up items that they don't recognize" via speech recognition technology instead of scanning or manually entering a price look-up code, he added.

Developmental teams at Balls will work with engineers at the Human Interface Technology Center of NCR, formerly AT&T Global Information Solutions, Atlanta, in the "user-centered design" project.

"That means the users actually do the design work," Riffel noted. "That's why we don't know how this will evolve; it's very much a 'bottom-up' strategy and works quite well. It ensures the success of the project."

Balls had already established various "development centers" throughout the store that test and evaluate systems. The retailer's front-end and produce development teams have been assigned to this project.

Riffel said although design has not yet begun, the broad concept calls for a computer-based system to respond to a wide range of spoken words and various dialects from around the country.

Cashiers, fitted with head-mounted microphones, could speak the word "apple" to prompt the computer to display color images and full names of various types of apples.

The immediate application will be for training purposes, Riffel said, but later could be incorporated into front-end operations.

For example, once an item is identified by matching it to the computer image, the cashier can speak the full name, "red delicious apple" to prompt the point-of-sale system to record the purchase without using a scanner or keyboard.

Should Balls ultimately introduce the technology to the front end, shopper's perception should be positive, Riffel said.

"The shopper would know the cashier was not making a guess, and we would not have to send somebody back to produce" to check an item's price, he added.

Riffel said produce misidentification does not always result in increased shrink; rather, it raises questions about price accuracy. For example, the retailer found, through the use of mystery shoppers, that emerald mandarins were rung up incorrectly as limes 90% of the time, he added.

"The problem is, the customer gets the register tape home, sees that they were charged for limes when what they were really after were emerald mandarins, so it becomes a customer confidence issue," he said.