[Indic-computing-users] Vernacular SMS on its way

[Ashish K. <as...@mi...>]

Vernacular SMS on its way
Vivek Shankar, ZDNet India,
March 04, 2003
(http://www.zdnetindia.com/news/features/stories/77297.html)

 11 million mobile phone users SMS primarily in English. A Canadian company,
Zi Corporation, has introduced predictive text input for SMS in Hindi.

According to the Cellular Operators Association of India, COAI, the country
had 11,163,141 mobile phone users in January 2003. That's a healthy 6.5 per
cent increase from the December '02 figure. The COAI does not give any
statistics relating to SMS or text messages sent by these 11 million users.
Nevertheless, it's a given that the number of text messages sent is mammoth
figure.

Taking a cue from a mobile carrier: how about adding SMS to the list of
"roti, kapda, makan aur Mobile"? For the moment, however, the millions of
messages sent by Indians have largely been in the English alphabet. And it's
an acknowledged fact that the English-speaking are a minority in India.
Ironically, a firm from Canada, a bilingual country using English and
French, is bringing SMS to the masses in India.

Zi Corporation has announced what it claims to be the first predictive text
input (also referred to as dictionary mode) for mobile phones in Hindi. The
Devnagari script is complex unlike the English alphabet. Hindi alone has 33
consonants and 11 vowels. Add ligature - combinations of letters to form new
letters of new shape, to these 44 characters and you have a horrid
experience of creating text; even with a keyboard equipped with more than
one hundred keys.

eZiText Hindi, Zi's offering, uses an innovative keyboard layout - vowels
are clustered on the 3 and 9 keys of the keypad. The predictive text input
is similar to that used in phones today; multiple key taps are not necessary
for character selection. Company officials also claim that once a user
inputs a word, it is predicted faster the next time. Trevor Sponagle,
Product Marketing Manager, told ZDNet India, "Zi has designed a learning
feature we call the Used Word Dictionary (UWD). The UWD stores words the
user employs often and predicts them faster during the next time they
write."

The software includes 'personal vocabulary', which predicts customised words
from a personal dictionary. Responding to a question on how much this
feature depends on the memory of a handset Sponagle said, "Yes, this feature
employs device memory; but very efficiently. Devices can store 100''s of
Hindi words for several KB. This feature is scalable for the manufacturer."

Zi was recently awarded a Chinese patent for its eZiText Chinese offering.
Commenting on the difference between Mandarin text input and Hindi input
Sponagle averred, "Our eZiText Hindi solution is very different from eZiText
Chinese. Chinese is an ideograghic language comprised of characters.
Characters can themselves comprise a word or combine with other characters
to form a word. For Chinese input, our solution is based on either
''building'' the character with stroke elements, or inputting the phonetic
pronunication, then converting to characters. From our beginnings with this
very difficult language, we developed solutions for alphabetic languages
like English. We now offer solutions for 100 alphabetic languages. Hindi is
similar to these solutions, except for the keypad layout - Hindi has more
letters - and the behaviour of ligatures letters joining to form shapes."

Another plus for eZiText is that it does not limit the user to a single
language. Sponagle clarified that it uses the same core engine for all
alphabetical languages. Accordingly, most eZiText implementations have at
least two languages, generally a regional language, say Hindi, and English.
He added, "Language input in the same message can be mixed, as well."

Regional language text input is handset dependent because the input software
is embedded in the handset. Sponagle said that carriers also play a part by
specifying that manufacturers include predictive text input for regional
languages in their products.

Zi also offers another product - eZiTap, a predictive multi-tap system,
which is language independent. eZiTap has a predictive engine but no word
database. The engine 'learns' words, which are multi-tapped the first time,
and predicts these words later on. This offering has the potential to take
SMS to every dialect spoken in India. However, one must note that a handset
needs to support the font/script of the language in question.

eZiText Hindi was launched in January this year and answering to a question
on the possibility of eZiText in other Indian languages Sponagle said, "We
develop languages based on our licensee demand; so we expect to provide more
solutions as the Indian mobile market grows and manufacturers see increased
demand for more localised handsets."

Hindi SMS should be a reality very soon - according to the company - Zi
licensees have already embedded eZiText for upcoming India models. Sponagle
was optimistic, "We hope to see the phones launched very soon."

A flash demo of eZiText Hindi is available at www.zicorp.com

===========================================================
Ashish Kotamkar (as...@mi...)
Mithi Software Technologies Pvt. Ltd., Pune. India.
Web: http://www.mithi.com
===========================================================
Communicate in your own language. Log onto www.mailjol.com.
===========================================================