ITyX and Fraunhofer Institute develop new Technologies for self-learning Data Extraction

Posted by Andreas Klug on May 12, 2010, 6:04:00 AM
Andreas Klug

New self-learning processes improve the recognition rate of required information from letter mailings and accelerate incoming mail processing.

 

The Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS and ITyX have developed a solution for the automatic extraction of business data. Business data includes, for example, customer numbers, but also address data or product information that is required when processing written enquiries.

Using this solution, an organisation that checks its incoming mail can now see if and where a mailing contains the required business data. Up to now, solutions were not able to differentiate information on their own without rule types and additional filters; for example what is a customer number or a similar looking telephone number? In a worst case scenario, incorrectly recognised business data led to expensive and time intensive post-processing. "With the new intelligent processes, the rate for business data extraction - in particular with free text - can be increased considerably, without having to develop countless rule types," emphasises ITyX managing director Suleyman Arayan. Customers will be delighted as their written enquiries are now processed much faster.

"The extension of the established ITyX solution shows how open minded SMEs can technologically round off their own products, without having to develop these time and cost intensive solutions themselves," said Dr. Melanie Gnasa, head of the Text Mining group at Fraunhofer IAIS. "The long-standing cooperation between Fraunhofer IAIS and ITyX also confirms how productively applied research and practical experience can work for business," adds Arayan. In its future cooperation, IAIS and ITyX will continue to place great importance in identifying new company specific problems that occur when processing incoming mail and to provide practical solutions based on research findings.

 

Topics: 2010