"We can build workflow applications for e-governance"

Peter Hauser is very clear about the potential speech recognition
technologies hold. In his recent visit to the country, he outlines the key
verticals that Scan Soft intends targeting through its alliance with HCL
Infosystems. An MBA from Boston University, Peter is keen about developing
solutions that will address the needs of local market. He feels that future
technologies will  be speech-enabled. In an interview with DQCI, Peter
talks about the  company’s channel strategy and plans for investment i
the country.

What is the nature of your tie-up with  HCL Infosystems?

The relationship with HCL Infosystems comes close on the heels of our
acquisition of Lernout and Hauspie. And we are looking at strengthening it
further. Our tie-up with HCL now extends to ten countries spanning the
south-east Asian market. It will distribute the entire range of digital imaging
products that include OCR and speech recognition software. Under the speech
recognition portfolio, we are also looking at HCL making inroads in the telecom
and the embedded market.

When you say embedded market, what type of applications are you talking
about?

Intelligent devices that respond to voice commands is what will drive the
market in the future. For instance there are more applications for speech
technology on mobile phones - SMS dictation, information access and so on.

Similarly the automobile market has a lot of potential for instant navigation
and various similar functions of a car can be voice-activated.

Has speech recognition evolved to the extent that a critical application
like navigation can be driven by it?

Speech recognition technology has evolved over the last few years and is
very reliable today. This is also because of greater processing power in PC and
even in non-PC devices. We have now built sound samples into the dictionary that
improve with use.

For instance Dragon NaturallySpeaking 6 comes with a vocabulary of 2.5 lakh
words. It also comes with a special filter to take out pauses during the
dictation. And an hour and a half of training is sufficient for the user to
start using the new product that will then give as much as 98 percent accuracy.

Does that mean that customers upgrading to Dragon NaturallySpeaking 6 will
also have to train the software?

It is not mandatory for the upgrade customers to train the software again,
but we recommend a re-training. This is because version 6 of the software has a
much bigger vocabulary and this will only lead to greater accuracy.

What will be your strategy towards addressing the Indian market?

We have crafted a well-organized approach to address the Indian market. We
will also look at developing solutions to address the local requirements. We
will target verticals and are developing vocabulary that is typically used only
in segments like finance, legal, healthcare, automobile, telephony and so on.

We have inherited the requisite reach across the length and breadth of the
country through the HCL channel. So in a way we would be able to ensure that
ScanSoft products are available throughout the country.

What can the channel expect from ScanSoft?

We are looking at two kinds of partners. The first category will comprise
those who have a retail presence.

The other category of partners will primarily be value-added resellers or
VARs. These partners must have the potential to deliver solutions. We will work
very closely with them and also provide them with the requisite training and
consultancy. And that precisely is the market that we wish to tap through these
relationships. We are also looking at a VAR certification program.

Will you recruit a new channel or let HCL develop the channel for your
products?

Since HCL Infosystems will be the ScanSoft face here in the country, it will
be HCL that will do partner recruitment.

All our channel initiatives will be driven by HCL Infosystems. However we
will work closely with HCL to see that we have a presence in all A, B and
C-class cities.

Are you also exploring the possibility of introducing local language
speech recognition software?

Adding regional language features is a constant exercise for us. We already
have software that recognizes Japanese, Korean and two versions of the Chinese
language.

Our core business is to develop and expand our language factory. We have set
our goal to achieve certain language milestones. We constantly undertake studies
to calculate how much time it will take to develop a language capability.

Similarly we have already initiated a market study to assess which local
language should be our top priority, should it be Hindi or any other language?
Also we have to look at the commercial aspect. We need to answer the question
whether there exists a sufficiently big market to come out with a product.

Does this imply that you have the capability to develop software to
recognize any language?

Yes, technically it is possible for us to develop recognition software for
any language. And in certain instances, where commercial viability is a problem,
the government has come forward to pool in. The Malaysian government and the
local government of the Basque region in Spain have funded such projects. This
helps the government to ensure equal accessibility to information for all.

Are you also looking at working with the Indian Government on such
projects?

Yes, we will look at developing applications that address the needs of the
government. With capabilities in character recognition and speech recognition,
we can build workflow applications that are suited for e-governance.

What then are your plans for investing here in India?

We are in the process of jointly building business plans with HCL
Infosystems. Both organizations have resources and we have joint plans for
investment too.

It would be too early to comment because we are waiting for the plans to be
formalized. I can however say that we will go beyond the standard methods. And
our investment in both research and development and in the market will largely
depend on the outcome of these plans.

MOHIT CHHABRA in New Delhi