for inviting me.

It's an honor to be here.

Trans Flick is a

third generation of machine translation technology that was originated in New York by the
Rockefeller Foundation when they hired a scientist to use machine language to translate Russian
to English using very primitive computer pre computer technology.

The second generation started with the ruled based translation the 1960's the CIA financing a company called
Cistern, that eventually moved to France and that technology now is the brand name in the
industry, but it's still the first,

second generation technology. The third generation is statistical machine translation
exemplified by a company financed by Incutel called Language Weaver which is a
statistical machine translation. The fourth generation is hybrid of both ruled base and statistical machine translation
and the next generation is incorporating human quality translation memory which is our
level. So we're taking best read approach in corporate hybrids of existing second, third, fourth
generation machine translation and we're adding human quality translation. Our innovation is extending
that best of breed platform to mobile and integrating it with voice for voice translation. And
customize linguistic data similar to a spell checker where you would add your words to a spell checker and the quality of
spelling goes up as you accumulate linguistic data that's appropriate to your subject domain. So we
now have 157 subject domain dictionaries in 16 language pairs. Our goal is

to increase from under 10 million words under management to 500 million words under
management in the next few years and added language pairs as they become more mature. So now
we're negotiating Bengali, Hindi, Ba Sa, Indonesian, Wardo. Some of which are coming off the
assembly line for military applications. We're a black light partner which already roams over
seventy carriers in over a hundred nations.

China being the largest market with 28 twenty percent of all handsets are now in China. The pie chart you see
to the right has changed a lot in the last 10 years. It was ninety percent English in 1995. It's now 30, 35 percent English. Chinese is number 2 with12.2 % , followed by Japanese, Spanish, German and
Korean. English will decline to 10 % of Internet traffic over the next I would say the crossover
would be in about six years at current rates at a 150 to 175 million people going
on the both the mobile and fixed line Internet every year who do not speak English as a first language.


So, there is now a language barrier is going online. There's linguistic isolation as a result. People have
to communicate in their second language of broken English. Some of which goes from fluent
English too limiting, limited English proficiency to know English. But when you get into terminology rich global
collaboration it's important to use a dictionary just like using the right spell checker with your words.


When you add voice over IP,

or for example to an instant messenger, it's important to know whether not
somebody has your terminology when they are speaking to you. Even if they speak 800 words
of English in common when you get into technical negotiation or business negotiation, you may
run out of words and have to switch to an instant messenger that has real time language translation. So we're
now going to be launched in February in the scight platform, 52 million registered users in scight, 190
million downloads 90 % non English speakers many of whom speak English as a Second
Language. But our so called broken

machine translation can actually be an improvement over somebody's second language of English if you add the dictionary.
We all know that

once you get the noun slot right, you get grammar and syntax 80 to 90 % right in Style, 95 %
right, you can be Republican president of the United States.

Meaning you can function with the noun slot as long as the noun slot is context sensitive. So what we do is customize
the linguistic database and make this go from get the just quality to something which is actually usable. And Bill Gates himself
has said machine translation becomes usable for Microsoft once we domain specificity. So here's our
interface which is patented. We also won a patent on dictionary customization over wireless network
So you select the service SMS, there's a billion SMS a day trying a billion and a half SMS a day. Now in Europe,
both becoming multilingual globalized networked societies. You can then select the language pair, say
English to Chinese, English to French, and the third slide you will be able to select a dictionary
including your idiomatic expressions, your jargon, your industry terminology.

Add voice, imported to voice mail, now you have voice to text plus machine translation which is customized
horded to voice mail so that 28,000 thousand teachers in New York can communicate with 600,000
parents who speak English, as a second language or no English and don't communicate
well. so the child is flunking. Jose is flunking algebra. It's important to know when the homework is do
so that you can ask Jose to stay home until he finishes the homework. If this conversation can occur across language borders,
even within our cities that are multilingual that could improve test performance. The same is true for Homeland
security collaborations say between U.S. and Mexico. A lot of these conversations that must take place
on immigration border control, immigration issues, security issues should take
place, but they don't take place because there's not enough simultaneous interpreters sitting at the desk of the staff member in the department of
Homeland Security. So over secure network, you can now have collaboration on a Blackberry or any major handset. The revenue
model is for 3 dollars a month flat rate.

10 to 15 dollars a month for customized quality with 16 language pairs seamlessly integrated with
SMS mobile email, mobile instant messaging over Qualcum platform which is Brew but which is a
derivative of Java. Java for Sun which is J2ME to me on a mobile handset or Simbian 7.0.
There's also Linux and Palm and IMODE in Japan.

It's client-server

technology so it's available https secure or HTML or over a SMS which is not over
the internet. And we add translation memory which is available if it's not found in the machine translation
dictionary it can search in a data base which is a 100% accurate of fragments and complete sentences
. So the accuracy can actually improve when you go from raw machine translation quality which is what
everybody thinks of when they think of machine translation. When you customize the dictionaries,
that's what Bill Gates thinks of 80 to 90 percent accurate. And when you add translation memory you go to
95 % or higher when you stick to the topic. Once you go beyond petroleum engineering, the bit
changes from petroleum bit to say computer science bit to drilling bit or horse bit or bit into an
apple. So the word bit could have 8 different connotations depending on the dictionary, so if one wants
to now globalize communication and be able not only break down barriers using the Internet, but break
down the final barrier to globalization of E-Commerce, one needs real-time translation. By
the time you translate it, travelocity or the savored database into Chinese, the flight's already have taken off
two days ago. It takes a couple of days at 1200 words a day, same with Ebay 15% of Ebay is
cross border transactions. Why shouldn't a portion of those 15% cross border transactions be
translated with cheap second tier quality translation with a correct noun slot? That would increase the volume
global e-Commerce. So our goal is to translate a portion of those billion SMS a day in China. A portion of those SMS
a day in Europe and create pan European, pan Pacific, trans Pacific, trans Atlantic global

IP communication. And voiced translation in the next generation. We did it for the U.S. Army
with of real-time translation of Arabic after 9-11 with 38000 military terms in the noun slot. That's about
60 to 80 % accurate depending on user but, and the domain because the domain can change to
crisis management to humanitarian relief even inside the army. So we started with NATO terminology which
covers only a certain number of crisis management scenarios, but now we have to customize for
humanitarian assistance for the army. So, then we have to add the Haitian, Indonesians so they could help
people in that part of the world. So that's really a broad sweep of all the different technologies which
thanks to XML and SOAP, service oriented architecture can now be fully integrated seamlessly with
security features with customized terminology to actually make raw machine translation military
intelligence collaboration quality presidential quality. Thank you.