BGU scientist produces first-ever real-time Arabic writing program

System lets user write Arabic words by hand on electronic screen which analyzes text and translates it into printed Arabic.

By JUDY SIEGEL-ITZKOVICH
herut arabic 298.88 (photo credit: Channel 10)
herut arabic 298.88
(photo credit: Channel 10)
Readers and writers of Arabic around the world will have Ben-Gurion University researchers to thank for creating the first software program of its kind in the world that identifies Arabic handwriting in real time, with an error rate of less than three percent. The system enables the user to write Arabic words by hand on an electronic screen, which then analyzes the text and translates it into printed Arabic letters in a thousandth of a second. The Beersheba researchers from BGU's department of computer sciences, Dr. Jihad El-Sana and master's degree student Fadi Biadsa, developed the system on the basis of research that uses calculative geometry to analyze the graphs, forms and dots that build the hypothetical reference points. The system identifies average handwriting and is built to learn new types of handwriting through use. A mouse and keyboard are the most common devices used to pass textual information to a computer. Hand-held computers and devices that allow writing on a small screen by hand require new methods of handwriting identification and its translation to printed text on computer. "When computer use becomes an important element in giving service to citizens, there is an enormous advantage in using handwriting instead of typing, especially in developing countries," said El-Sana. "Today there already exists a system for identifying Latin handwriting in high percentages, but there is no system to identify Arabic handwriting, which is used by around one-sixth of the world's population. The identification of Arabic handwriting is difficult because it is joined and most of the Arabic letters contain dots either above or below the letter. It should be stressed that we have an obligation to maintain the same style of writing without change. For example, in some models of hand-held computers even in Latin letters there is a small change in the letters, so that it will be possible to identify them as fast as possible without errors." The target population for the Arabic handwriting recognition program is around a billion people who speak Arabic, Urdu and Kurdish. "As far as we are concerned, the sky's the limit," El-Sana said. B.G. Negev Technologies, the university's technology transfer company, is currently seeking investors for the project. El-Sana specializes in 3D interactive graphics, multi-resolution hierarchies, geometric modeling, computational geometry, virtual environments and distributed and scientific visualization. He received his bachelor's and master's of science degrees at BGU and his Ph.D. at the State University of New York at Stony Brook and has published over 30 papers in international conferences and journals.