gil@svax.cs.cornell.edu (Gil Neiger) (01/21/87)
For what it's worth, here are the results of my "survey": ========================================== From Abdelsalam Heddaya (Harvard University): I do know of two research projects in Egypt which, if successful, should satisfy your friend's requirements. At the U. of Alexandria CS Dept, a guy called Mohammad Siddiq is doing his Masters thesis on a system to recognize hand written Arabic characters. The main limitation of his work is that he could not (as of last summer) recognize but individual characters. This is a problem, since Arabic words consist of connected ligatures from a very rich set of shapes, while individual characters each has only one shape. Mohammad's supervisor is Prof. Amin Shokry, and he has long been interested in this topic. If anyone knows of the existence of such a system, I believe he will. A similar project was in progress last summer at IBM Cairo Scientific Center (CSC). Osama Emam, who is a junior researcher there, was building a system to recognize printed Arabic words (as opposed to individual characters in Siddiq's work). Also at IBM CSC, there was a project to develop a program to analyze Arabic words morphologically. This means reducing word to its root, and showing how it was derived from thatroot. I believe such a tool would certainly have a spelling correction capability. Unfortunately, I don't know the extent of completion or success of these projects, but if your friend is interested in experimental systems, he can contact the above mentioned people. Here are the addresses: Mohammad Siddiq (or Prof. Amin Shokry) Dept. of Computer Science Univ. of Alexandria Alexandria, EGYPT Osama Emam (or Dr. Mohammad Hashish, the director) IBM Cairo Scientific Center 56 Gaami'at al-Duwal al-'Arabiya Street, 12th floor al-Muhandeseen, Giza EGYPT In addition, Osama has a VNET address which I never succeeded in using. It is: 91425595@UITHONE. Good luck to your friend in his hunt, and I'd like to know the results. ========================================== From Pavel Curtis (Xerox PARC): The Xerox Viewpoint software can handle Arabic and recognises words, etc. It can also mix English and Arabic. I don't believe that there is any light-pen style interface to it, however. ========================================== From Todd Olson (Cornell University): Well, I think it was in ABACUS (the comp mag from springer-verlag) (I don't recall the issue but since there are only 4 a year ..) that I saw an article on an arabic system being developed for some country in the mideast. It was supposed to be working on something small. May have been a IBMPC. (This is a vague pointer. I know saw such an article and I'm pretty sure I didn't see it in 'Computer Language' or Dr Dobb's but it may have been in Scientific American.) ========================================== From Betty Lou McClanahan (MIT): I recall seeing an article in Aramco magazine some time ago about Professor Abboud at the University of Texas (at Austin?) teaching Arabic using a computer program. If I think of it, I will try to check on it. - Betty Lou McClanahan (another friend of Amr Abbadi) ========================================== From Michael Lee Gleicher (Duke University): The computer assisted language instruction project at Duke has done programs for PC's in many foreign character sets. Last year I developed a bitmapped graphics library for them so they could do Amheric and Modern Greek, I think they were going to Arabic too. You can write to: Jeff Gillete Department of Germanic Languages Duke Univ. Durham, NC 27706 I think you can reach him as duke!ducall!jeff also. ========================================== From "Hashem" (communicated by Abdelsalam Heddaya): Here is one person you may contact about a program for parsing Arabic and for deriving words automatically John McCarthy Phone (201)582-6595 ========================================== I checked out the July-August 1983 issue of Arab World (produced by Aramco), and found the article about Victorine Abboud at the University of Texas at Austin (not to be confused with her husband, Peter Abboud, the author of Modern Standard Arabic). There she developed, while working on her Ph.D. thesis in linguistics, a program for teaching Arabic. It seems quite interesting, and a good piece of work. It consists of programmed lessons, and is allegedly very sensitive to the progress of individual students. It teaches and tests Arabic writing, although it is not clear from the article whether or not it can "read" the writing of the students. Apparently it works with a bilingual keyboard. The article (which is almost four years old!) says that Mrs. Abboud wants to transfer the "program from the university's computer system to an IBM personal computer that has a special graphics board providing a resolution of 700 horizontal by 350 vertical dots per inch." The article refers to the program as "CAI" - computer aided instruction. It mentions no other name for it. ========================================== Thanks to everyone who helped! - Gil