--
Regards
-----------
Sachin Rase
MediaIBC 26178979
----------
--- Begin Message ---Hello Mr BGanesh and All, Here is a short(!?) summary of our IndLinuX Team work:Today the TeNeT Group (IIT-Madras) is involved in the wide scale deployment of Internet kiosks across the countryside. The kiosks require local language software so that they can be effectively used by the marginalized literate population of OUR country. Given the number of Indian languages and different encoding standards and fonts they use, it was felt that we need to develop Indian language support that is essentially INDEPENDENT OF ENCODING & FONT. Further, we felt that this support must be provided at the kernel / X-Window system levels so that applications that run on top of the kernel/X-Window system INHERIT the interface. To this end we provide TWO different solutions under GNU Linux: kernel based and X-Window based. > Kernel based solution: Indian languages have variable width characters. To support variable width at the kernel level, a multi-glyph approach is proposed, i.e. a keystroke may correspond to one or more glyphs. Further, the number of keystrokes possible on the standard QWERTY keyboard is not sufficient to represent all characters and character clusters in any given language. This results in the problem that a sequences of keys may correspond to different glyphs (in that a key sequence AB might correspond to a glyph G1, while a key sequence ABC might correspond to a different glyph G2). This requires that the system has a memory of the keystrokes that have been made. To address this problem a finite state machine and a set of associated rules was introduced in the kernel. > X-Window based solution: Two different solutions are supported at the X-Window level, namely, glyph based and encoding based. In the first approach, similar to that of the kernel, the parsing is done on the input side, while glyphs are produced on the output side (file/display). In this approach, clearly the system is tied to the font that is used. To overcome this problem, an encoding based solution has also been provided. The problem with this approach is in cursor positioning. The position of the cursor is maintained by the application which interacts with X to obtain the widths of characters. The cursor is positioned based on the keys typed (no memory/context sensitivity is assumed). We are yet to address this problem. We do feel that the only solution available to us is to provide a set of cursor APIs in X which can be used by any application. In either solution, the language dependent information is provided in parse rule tables. This information is loaded at runtime. This ensures that changing the local language environment is SIMPLY a question of manipulating parse rule tables. > Applications: The solutions that we have provided enable a user to type in a local language in most applications (Open Office/ KOffice/gcc...). We have further customized some of the applications (menus/errors in local languages), namely, KOffice, Konqueror, gcc, pine, emacs. We are currently working on customising Open Office. We are also working on a convertor which converts indian language documents from fonts to encoding, fonts to fonts, encoding to font and it can also do translitration across indian languages > Current Support: Tamil (Glyph-based) - TamilNet99, Inscript Hindi (Glyph-based) - Inscript Tamil (Encoding-based) - ISCII Hindi (Encoding-based) - ISCII The above supports are complete (This simply means that we have populated the appropriate language dependent parse rule files). We are currently working on Malayalam, Oriya, Marathi and Gujarathi. > Major Bugs o Cursor position problem o Few Memory leakages > Future Plans: This work is part of a MAJOR PLAN to develop a MULTI-MODAL interface to the computer. By MULTI-MODAL, we mean, handwriting, keyboard and speech interfaces. We have started work on both handwriting and speech. We have developed a handwriting interface for Tamil and have developed speech technology software for: Speech synthesis in Tamil, Hindi, Telugu and Malayalam (Malayalam database was obtained from ISDL, Trivandrum). Spoken language Identification, Syllable-based recognition of continuous speech, Speaker identification. Web Site : http://www.tenet.res.in/Donlab/Indlinux/ ( Currently reconstruction is going on) Send ur Comments and Suggestions : [EMAIL PROTECTED] Thanks, --- Bharathi S, IndLinuX Team, (__) DONLab, TeNeT Group, oo / IIT-Madras, Chennai-INDIA. (_/\ _______________________________________________ To unsubscribe, email [EMAIL PROTECTED] with "unsubscribe <password> address" in the subject or body of the message. http://www.aero.iitm.ernet.in/mailman/listinfo/ilugc--- End Message ---
