The Hindu Business Line : Making a difference, and making money

You feel grateful to board the air-conditioned bus to escape Chennai’s sweltering heat. And this is no ordinary bus. Google has designed it to showcase all that can be done with the Net today, hoping that as it travels around the country, the bus will introduce the magic of the Internet to countless people.

As you forget the heat, Prem Ramaswami, product manager at Google, takes you through the story of the bus and how he, having spent most of his life in the US, is passionate about making a difference here in India.

After reluctantly leaving the bus, you are ushered into, thankfully, an air-conditioned tent, to meet with Dr Prasad Ram, Head R&D, Google India.

And you hear him start on similar lines — about making a difference. Incidentally, Ram was at the Pan-IIT meet in the city about a couple of months earlier. He talked as intensely about the experience of a lady, Jasudha Ben in Gujarat, wanting to use the Internet so that her son could join the mainstream.

Then you wonder: what is it about Google that has its employees talking in this strain? Isn’t the profit motive top of mind for all employees? Does only someone at the very top need to worry about it?

Ask this of Ram and here’s what he has to say:

If I can eliminate information asymmetry and give everyone access to information, their livelihood will improve, their lifestyle will change and productivity will go up.

If you want to select a school for your child, you currently choose based on reputation, recommendation and brand recognition. If you have access to right information such as size of class, student-to -teacher ratio, extra curricular activity, and other relevant statistics, it will drive the right choice. It will incentivise the right behaviour from the school so that they deliver to those metrics.

If I want to run a disaster-relief operation, say following an earthquake in Kashmir, I don’t know where the hospitals are or where the exit routes are. I don’t know how much load those routes can take — for supplies, trucks, troops…

Currently, after the earthquake, we’ll start figuring out some data and come up with a management plan. But the right way is to start with the data.

Assume you are a farmer wishing to sell (sapotas) chikoos in the Hubli market. First off, if you have information on what other farmers have planted, you wouldn’t plant chikoos. After harvesting, if you know that prices are better in the area five km westwards, rather than five km southwards, you would go to the right place.

This is the power of information. When you overlay your company’s mission to organise all of world’s information and make it universally accessible and useful, then you do it because it’s good for people and good for profits. This is the golden mean — it doesn’t get better than that.

The passion to make it happen comes from the wide impact it could have and the profits come from the advertisements that this would attract.

Internet penetration

If you compare the PC penetration scene now versus 15 years ago, it is better now. But even then it could be even better. We only have 45 million Net users in India. It forces me to introspect and ask so why aren’t we using this?

To go back to the story of Jasudha Ben — I actually stood outside the Internet bus in rural Gujarat, offering a Rs 20 note to anyone who would go in and browse the Net. I thought there would be this flood, and I had a wad of notes. Not one person took it. Then this lady, Jasodha Ben said “I have tried it once and the Net is not relevant to me.”

Look at the insights she gives: “I know Gujarati and your content is in English. I am a mobile user and you have desktop solutions. I want to search for hand-pump repair shops in my locality and you show me pizza stores in Ahmedabad!”

Our obsession is ‘urban’, ‘city’... She says that when she wants a bus or train route from Surat to Bhuj, she only gets driving directions. “What can I do with that, or with flight status?”

Relevance issue

The challenge for us is how we make the Net relevant to these people. And to really understand what relevance means. You would be hooked to the Net only if you can find content that is local to you. There is a global world and all that but there is a local world — I go to the same cinema, same kirana store, same tailor shop, same restaurant — all within a one-km radius of where I live. Just because I don’t live in Chennai or Ahmedabad but only in Krishnagiri, I can’t be out of the loop. Then the Net is not relevant to me.

Two, (as a search engine provider), I need a lot of content to drive meaningful traffic to my site. A typical store has only 5 per cent of its products contributing to 99 per cent of sales. I can’t stock only those products making up the 5 per cent, willing to sacrifice 1 per cent of sales. Then, people won’t come to my store because they come believing they would get anything they want here.

Likewise, we all go to the ‘English’ Net, because we have this great belief, that we will find whatever we want on the Net. It could be information on prostate cancer or trekking in the Himalayas…If you are a Tamil or a Gujarati speaker, you don’t have that comfort.

To me, the digital divide is essentially a relevance issue, not an affordability issue. If a mobile phone is relevant, a rickshaw driver would want it as much as you would want it. Price is not the issue — market dynamics would take care of that. If it is relevant to you, manufacturers would find out a way to get it into your hands — some plan or the other to make it affordable.

Improving relevance, with local information, eliminating access barriers such as language, form factor (mobile vs PC) barriers, is key. If you ask a mobile user to sit in front of a desktop, usability is a barrier. The metaphors we use in the desktop world, such as the file folder metaphor, are usability barriers. What if you have never sat at a desk in your job? You can’t identify with it. We have to address those.

There are a billion people or more on the Net. For those of us who are, we can swear that this is the best thing that has happened to it and we can’t live without it. The other 5 billion don’t want to get onto it. If it were so good, they ought to be on it as well.

Relevance is lost. (The have-nots) don’t say, “I too want that (be a part of the Net).” Our R&D challenge is to increase relevance, build products and technology that bring online more content, eliminate access barriers — usability, language, mobile.

Google India’s work

We have been working on projects where information is available but it’s a great challenge taking it online. For example, how do you take online 200 years’ worth of newspaper content? Chandrayaan happened recently. If you are a lunar enthusiast and want a full chronology of events that happened since man first went to the moon; or if you are a history student and want all details about the 1857 Indian War for independence; newspaper content would be valuable there.

Interested in knowing what else Google India has done? Logon to our Web edition.

We needed to build extraordinary image processing technology, which, unlike a book, does not go page-wise, but column-wise. The human eye and brain are great at this — finding out which three columns a particular headline is meant for; and to differentiate between classifieds and news.

We took all this, content from 1857, did image processing, extracted all articles, did Optical Character Recognition (OCR). Interestingly, acting on a search query, when we searched, we did full-text index, but when we show results, we show the newspaper.

Lack of data

In cases where information is lacking or does not exist as structure data, we build applications to make meaningful search possible there. Map Maker is an example. One of my engineers said, “There is no good content in India because a lot of this content is geo-referenced. It means most locations are referenced by the latitude and longitude at which they are.”

We also observed that there is no business incentive for anyone to put up maps. I can’t sell my maps data, and the like.

Then we came up with the idea of doing it on a Wiki model. Everyone knows the road they live in, the school they went to and the office they work in. Everyone can write that much on the maps… and then we can stitch it all together. So the whole nation paints on one single canvass. The technology is extra-ordinarily complex, resulting in the Map Maker last year. It was launched in 160-odd countries. India was the 57th in line, though the inspiration came from India and is strategic to the country.

If you search for a paan shop — someone was passionate about it and he marked it on the map. You go there and this paan shop suddenly sees more footfalls without the shop owner even being aware of this.

Language processing

Then there is this whole issue of language access — only 7 per cent of Indians are proficient in English. We are part of that and interact only with that 7 per cent. We are consumed with the confidence that everyone in India speaks English. All keyboards you see — and I have travelled to 100 villages in India — whether in the North-East or Bihar or the South are all English keyboards.

So we have to come up with a technology that helps users input information through English keyboards, in phonetically equivalent English — ‘mera bhaarat mahaan’ is an example. You write it out in English with transliteration. And, we don’t want users to learn rules. Type two a’s, or single h… whatever. We wanted to let the system figure it out, so that there is no barrier to doing this.

And we had to deal with more than one language in India, so I couldn’t bring in language-based rules. That meant building a mechanism that is an inherently learning mechanism. If you feed it English words and corresponding Hindi words, it learns Hindi transliteration. Likewise, for any language. The underlying product is the same and differs only on the data it is fed. I can do Manipuri transliteration quickly if I wish to, now.

This is a uniquely Indian problem — of 22 official languages. The opportunity is good since it makes us think systemically. I can’t have a short cut and attack one language. It enforces discipline. It’s the same with translation. I have to do machine translation. If I can get all English content in Gujarati — then the job’s done. For, no one is going to write about diabetes or prostate cancer in Gujarati for me in the near future. But there are tonnes of information on this from the National Institute of Health.

All I need to give the user is a button-click on translation. The technology allows the engine to learn translation. Once it does it for any one language, it would do any translation well.

It’s a piece of technology that is extremely complex and there is a lot of participation from Google centres across the world. But because of the strong need in India, we contributed significantly. It’s not fully baked in India.

Ram’s mandate within Google

I look at my job as three things:

What kind of engineering output are we producing? Am I building really cutting-edge technology and products? I don’t want to lose that. That is one important facet to me – trained an engineer, I will always be an engineer.

The India team is doing a huge amount of work for Search. In India, if you search for BSNL, you expect search results to throw up MTNL as well. In your mind, they are the same. GoI is obviously Government of India, but only to an Indian is it so obvious. The Search engine ought to be sensitised to that. Search for Hoysala, which is not an English word, and misspell it. You would still expect the search engine to show you the right documents.

We are a search company and want to ensure that that part is absolutely well taken care of. The user’s view point is “Show me what I need and not what I ask.” We are continuously at it but so far we have done a fairly spectacular job. It’s evident from people’s reliance on Google search for information.

The second metric is how we are doing in India. And, Google is doing remarkably well in India. Anecdotally you would know — and I know based on data. Google, Maps, Orkut, all of these are top notch products. And we have a huge role to play there and we are doing well there.

The third metric for my performance is whether we are building a great product with an impact worldwide for users as well on the technology dimension. News Archive, English OCR, Map Maker, ….The idea is to build global products with technology innovation. That’s what makes it work for us.

Acquisitions in India?

For us, an acquisition is driven by two considerations — great technology; and a good initial, start-up team, in that sequence. Sometimes some element of the first is there but the team is not as strong as we would like it to be. Generally, the first metric fails.

For example, if an Indian company had cracked Indic OCR, then I can OCR any product and voice-recognise any Indian language. But no one has done that. It would be valuable if someone had figured out a technology to recognise an address any which way we write — whether it’s the accurate address or just one line along with “Opposite Ganesha temple”. That’s not happened.

We have made investments in Erasmic Ventures, Seed Funds, TeNet, etc. But the idea there is to encourage entrepreneurship in the country. An acquisition is a different thing. Investment in funds is about being of value to the community from which we have derived value.

bharatk@thehindu.co.in