News Release

Researchers to convene to explore how computers can better read handwriting, music, even sanskrit

Meeting Announcement

University at Buffalo

New ways of bridging the gap between the written word and the ability to access it electronically will be on the agenda at the fifth International Conference on Document Analysis and Recognition (ICDAR) to be held in Bangalore, India on Sept. 20-22.

More than three hundred researchers are expected to participate in the conference, which is being sponsored by the University at Buffalo, Pennsylvania State University, Microsoft, IBM, Siemens GmbH, Panasonic and others.

Topics to be discussed include new technologies that can be used in applications ranging from entering data into hand-held Personal Digital Assistants and 'reading' Chinese and Japanese mailing addresses, to understanding music scores, or preserving ancient Sanskrit manuscripts by storing them electronically.

Chaired by Sargur Srihari, Ph.D., SUNY Distinguished Professor in the UB Department of Computer Science and Engineering, the biennial conference is the world's largest in the field of document analysis and recognition.

"This conference is the world's leading forum for exchanging ideas about how to make the vast array of documents available in the world -- from airline tickets, bank checks and mailing addresses to maps and symphonic scores -- easily readable and accessible by computers," said Srihari.

Srihari directs UB's Center of Excellence for Document Analysis and Recognition (CEDAR), the largest research center in the world devoted to developing new technologies that can recognize and read human handwriting. CEDAR was established in 1991 by an agreement between the UB and the U.S. Postal Service, which has to date provided more than $25 million to the center.

According to Srihari, electronic commerce has increased significantly the amount of mail in the world, as items purchased on the Internet are sent to consumers. Since 1997, the postal service has been using CEDAR's handwritten address interpretation system to automatically interpret handwritten addresses on envelopes, enabling it to save millions of dollars in annual processing costs.

However, Srihari explained, except for Australia and the United Kingdom, which are beginning to implement CEDAR's system in their own national postal services, such technologies have not been implemented anywhere else in the world.

In addition, he said, there is a great need for document-recognition techniques to reduce the amount of paperwork on people's desks, particularly in the Third World.

"Some countries, such as India, are becoming international centers for processing certain kinds of paperwork, such as stock certificates and airline-ticket coupons, and the need for automation in these areas is urgent," he said.

At the conference, researchers from UB and other institutions will address some of the recent attempts to automate these services in Chinese, Japanese and other languages.

Theoretical and fundamental bases that make these technologies possible also will be discussed.

For example, UB researchers will discuss their new technique that allows computer systems to read and interpret two handwritten numbers that are touching without having to technically isolate them, research that ultimately will enhance further the speed of handwritten-address interpretation.

In another paper, UB researchers have developed a unique definition for determining the similarity of two handwritten characters based on figuring out in what order the writer made his or her pen strokes.

Other papers will focus on automatic interpretation of handwritten chemical formulas, methods for matching street numbers in Japanese addresses, a postal system with a built-in learning capability and algorithms for reading Chinese characters.

###

Conference planning and organization was conducted under Srihari's supervision by CEDAR staff at UB.

Additional information on the conference is available on its Web site at http://www.cedar.buffalo.edu/icdar99.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.