HP Labs India

Smashing India's language barriers

ZDNet UK News
May 12, 2006

India is a unique country with unique communication problems – no surprise in a country with more than a thousand languages

There's nowhere remotely like India: a billion people, 1,652 languages, three-quarters of the workforce in agriculture but immense numbers of literate, technical graduates pouring out of universities — yet it lags behind its major competitor, China, due to poor infrastructure and incomprehensibly complex bureaucracy.

According to a team of HP engineers from Bangalore, over to meet UK journalists in Bristol as part of the celebrations around HP Labs' 40th birthday, technology can help, but only if everyone can interact with it. Only around 10 percent of Indians can conduct a transaction in English, which has limited the penetration of PCs to around 60 million. There's no further chance of growth unless that particular problem is solved, the Bangalore team claims.

Voice recognition and continuous handwriting recognition have both been tried, but neither is reliable enough for daily use in the community. Instead, HP India has developed the Gesture Keyboard, or GKB, a digitiser pad and stylus combo that works with Indic writing, one of the major families of local language. This includes Urdu, Gujarati and Hindi — there are 400 million Hindi speakers alone, an enormous group to isolate from IT or, if you prefer, an enormous new market more numerous than the European Union.

The key characteristic of this way of writing is that it has a number of base characters in an alphabet, but each may have one of a large set of vowel modifiers — matras — added. For example, Hindi has 36 consonants that can be modified by any of 12 matras, and almost every consonant can be bound to another. That leads to around 1,500 symbols. Trying to input these on an ordinary QWERTY-style keyboard results in huge numbers of keystrokes. Even expert typists find it difficult to master: someone with no exposure to technology can find it alien indeed.

The GKB reflects the way the script actually works. You take the stylus and tap one of the base characters from the set printed onto the tablet, and that enters the character. Or, draw the matra over the character on the tablet: the underlying gesture recognition software in the product then combines the modifier with the base character to produce the right result. HP says that because this is very similar to the way writing is taught in schools, it takes around twenty minutes to pick up and can soon produce around twenty words a minute.

Although the device is relatively expensive — it costs around 40 euros, in a country where a PC can be bought for 200 — HP says it can easily double the income of a cybercafe or kiosk, one of the burgeoning one-man band computer booths that give non-computer owners access to the Internet, government services and so on. It might also be built into keyboards, either by replacing the numeric keypad or by sliding out from beneath.

The GKB is currently available in Hindi and Kannada, the language of the province of HP Labs India, with other versions such as Persian and Arabic in the labs. It takes about a month to add a new language, providing it follows the same basic rules; the project itself took a year to produce, building on four years of work in general purpose handwriting recognition.

PrintCast is another example of technology being adapted to specific needs. This is a way of embedding documents and other data types in video broadcasts: in principle, nothing too different to the way Teletext and other systems have been sending information in TV signals since the 1970s, although PrintCast also understands digital transmissions. But by focusing on printable documents, PrintCast can massively improve retention of the information in the broadcast — very important in areas where access otherwise is limited. There are already plenty of satellite and terrestrial broadcast systems in India — the inventiveness comes in making them double as carriers of printed matter without modification. The new information is mixed in with the programme before transmission, and a separate decoder monitors the video signal at the receiver and copies the documents into a buffer, from which it can be printed.

The system is designed to be used in public forums: people come in to a hall to watch the broadcast, and at each point where a document can be printed, the person running the equipment can run off as many copies as needed, usually for a nominal charge. The advantages for health, agriculture and other local government information transmissions are huge, not least because in many areas other printed material distributed from central government is held up or not distributed at all — documents that tell people their rights are not always popular. This bypasses huge swathes of bureaucracy.

Neither of these two inventions use breakthrough technologies — in principle, either could have been built ten or fifteen years ago with small teams of engineers working with modest resources. Indeed, at the HP Labs Bristol demonstration, a number of journalists were disparaging about the systems — "So, you've plugged a printer into a television. So what?" But like all inventions, no matter how obvious, it took someone who knew of a particular need to make them happen — and their importance has little to do with the mechanics behind them.

The needs of the invisible world, the three billion outside the rich countries, are only just becoming known — as evidenced by HP Labs, which took 36 years to set up an Indian operation, and still got there ahead of most. With that, the company is betting that by meeting the country's needs the benefits will flow both ways. It's not just a matter of opening up another market, albeit one as big as any we've known before. By connecting these people, their ideas, capabilities and potential can flow back, bringing us new information and perspectives that can only help the innovation on which the industry depends.

Click here to read the article at the ZDNet Website

This page was last updated on March 2, 2010