What are the emerging applications? (HR or human resources, GIS or geographic information system, education, solutions for the physically challenged, speech biometrics, home, professional productivity tools such as ‘Journo’, music on demand?) Does speech technology permit the integration of other cutting edge technologies? How?

There is no limitation to application of speech technology. With changing times (mobile penetration), the most appropriate technology for mobiles is ASR. Speech is becoming a convergence tool, across many ‘day-to-day’ affairs of the common man. The next wave is location-based services (LBS) combined with ADA (automated directory assistance) or Yellow Pages. It is predicted to be about $10 billion by 2010.

Speech is now replacing keypads and keyboards (in future — it is predicted that mobiles would have only mic and speaker with optional display – keypads would vanish). LB would slowly — but steadily — make all relevant use of ASR applications to improve productivity and bring enterprise efficiency — resulting in growth of our customers — multi-fold.

Also, I see application of speech biometrics across multiple areas in customer interactions over phone as well as in internal applications such as the IT-helpdesk using the solution to provide various services efficiently.

LB would be concentrating on helping the physically challenged, as I am personally committed to doing this.

Any interesting insights (about human languages) that you have gathered over the years you and your team have worked on speech technology. (Wonder if speech systems work in non-human environments too!)

Handling “Madras Tamil” is the funniest experience. For most of our applications, we write a “grammar” for recognition. This completely fails with typical “Madras bashai”, as it is not Tamil in many ways. This made us adopt “mixed” language (Tanglish) for many applications. The non-human environment, we have not tested this application yet.

Where does ASR (automatic speech recognition) fail? Or, where is it not apt?

In a very noisy environment – say in a loom factory where noise is more than signal – ASR would fail. In India, we have the habit of honking even when not necessary and in such a place (say, the roads), ASR does not work very well. It does not require soundproof rooms, but relatively quiet places would give excellent results.

Again, when a person has a “grievance” and he/she wants to talk to a person, someone they believe to be trustworthy, it is better to have a human reassuring that the grievance would be redressed (at least an assurance). Putting an automated system in such a situation would only lead to further aggravation of the grievance. In some such cases, ASR is not the best fit. But, in more than 90 per cent of the situations, ASR would be more convenient and simple.

On the IP (intellectual property) creation at LB, and also the IIT advantage. And the skills you find most relevant to your work.

Currently we use COTS (commercial off-the-shelf) product from Nuance and do the “regionalisation” and “localisation” to suit an application and territory. But, to realise the dream of “bridging digital divide” – for a target of 700 million people to use ASR for transactions and information – we cannot be dependent on COTS product; it would be cost prohibitive.

LB is, therefore, working with IIT-Madras to develop our own engines for ASR and TTS in about 3-5 years time.

Working with IIT-Madras has been great, as they are not only helping in R&D (research and development) but also invested in the company with an aim to capitalise the tech deployment when ready to rural masses, to bridge digital divide.

Towards this, we have already started work and have gone through a couple of iterations of R&D plus testing over the past 18 months. We also have commissioned a full-fledged ‘language resources’ (LR) team that would build a corpus of transcribed data for a new acoustic model and a language model for various Indian languages. A team comprising computational linguists is working on this already, which would be the key IP of the company in the future.

Our company has filed three patents so far for very cutting-edge application of speech in the common man’s life. This would add to the IP of company in future and help in valuation and also capitalising huge biz opportunities. Two more patents are under processing.

Do you think speech recognition systems can help serve larger interests too: such as – better governance (by facilitating democracy?), education, ensuring the right fit (of students in study choices and employees in jobs), and counselling (therapeutic?)

This is a nice question — there is nothing that can stop these applications. I see that many initiatives for e-Governance come with self-service kiosks, without realising that more than 70 per cent of the population cannot read the screen. The purpose of sharing the info through self-service kiosk is to cut the middleman. But, in some places you can find the villagers paying a ‘smart guy’ money to find info using “kiosk”!

We are thinking “big”, but would act once we have our own engine to proliferate into larger apps. Already some initiatives related to education segment and employment etc. are being worked with telcos.

We are also betting big on “speech biometrics”, which can securely identify a person at the other end of the phone and help in various transactions, including financial, trading and so forth.

You would see LB very visible in fuelling/contributing to the growth of India to be a “superpower”.

On the origins: why speech as a passion?

I am a trained geophysicist from IIT-Roorkee — one among the best institutes for geosciences in the world. We worked on “signal processing” – mainly to understand all the signals given by the Earth (electrical, magnetic, seismic, acoustic etc.). But, somehow I knew I would not fit in a Government/ PSU (public sector undertaking) job. GSI (Geological Survey of India), ONGC (Oil and Natural Gas Corporation), CGWB (Central Ground water Board), and OIL (Oil India Ltd) were the mainstream recruiters at that time.

I searched for an alternative in the computer field, as we were used to writing a lot of long programs to process signals to find oil, minerals, water and so forth. In 1986, it was easy to get a job in the then emerging IT (information technology) field.

Thus I have been in IT for more than 21 years…I started as a technical specialist in X-Windows and moved to sales and marketing, to handling very large projects for my companies, which took me to various parts of the world.

Speech passion probably comes from my signal processing background. I was amazed to find the possibilities of applications for socio-economic and cultural as well as technical applications of speech.

I have personally given the first four years of my professional life (after M.Tech. – from 1986 to 1990) to rural upliftment/betterment when I worked on the application of technologies like GIS and image processing to help resolve local issues in Bahraich (most backward district of Eastern Uttar Pradesh), Deogarh and Dumka in Bihar, Bundelkhand in Madhya Pradesh etc. working closely with people, on income generation at the local level.

As a result, I could understand the local issues in rural India and firmly believed that technology and communication improvements are required to make the country develop into a superpower. Which explains why I always kept searching for relevant appropriate technologies, wherever I travelled across the globe (in work-related assignments, as I do even now).

One such finding, while on a trip to the US (in early 2001) was ‘automatic speech recognition’ application (called ‘Hey Anitha’). I realised that my search for the appropriate technology is over and that ASR is an ideal solution to “bridge the digital divide”, and relevant for country like India. Then I began doing a lot research to convince myself that the technology is mature enough to make the best use of.

When I decided to be on my own, I knew CRM (customer relationship management) is going to be the biggest growth area as product differentiation would vanish over time and only “innovative customer interactions management” would be a key determining factor for growth.

Hence, with conviction we ploughed resources and took some risks in this technology – a “reasonably new” one worldwide (the second wave had just started in 2001). God has been kind to support with me the right people, right customers, right investors and right use of tech (so far) for our growth (we grew 376 per cent on topline – from Rs 2.2 crore to 8.29 crore, last year and also declared a maiden dividend to investors).