Ajit Balakrishnan: Attending to the other atom

PEOPLETECH

At 5.30 in the morning of the 15th of a July, in a stretch of semi-arid land, a group of men watched a mighty explosion and then turned to listen as their leader intoned from the Bhagwad Gita: "If the radiance of a thousand suns were to burst at once into the sky, that would be like the splendour of the Mighty one ..."

The year was 1945, the location was New Mexico, the leader of these men was Robert Oppenheimer and the men were scientists of the Manhattan Project and they had just watched the first atomic explosion in the history of mankind.

Shortly afterwards, an atomic bomb, the outcome of this first explosion in New Mexico, was dropped over the Japanese cities of Hiroshima and Nagasaki, killing 120,000 people and indirectly another quarter million, bringing the Japanese to their knees and ending World War II in the Asian theatre.

Since then, scientists have been trying to harness the enormous power of the atom to produce something more useful to society, electric power. At one time it was believed that electricity produced from atomic energy would be so cheap that it would not be worthwhile to even meter it. While these dreams have faded a little, with oil prices skyrocketing past the $60 mark, interest in generating power from atomic energy has been renewed. The Indian government is in the middle of acrimonious negotiations to license technology for this from the United States.

But while smoke and dust swirls around this latest effort to harness atomic power to fire up our industrial growth, there is another battle shaping up. The outcome of this battle may matter more for the next stage of economic growth and it is based on another kind of atom""the atoms of language.

Think of the time your school teacher surprised you with the information that what you know as common salt is actually made up of an atom of a shiny metal called sodium combined with an atom of a greenish gas called chlorine to make sodium chloride and that all the substances that we see around us in the world are made up of varying combinations of a just a hundred atoms of elements like oxygen and nitrogen and carbon.

Mark Baker, a professor of linguistics at Rutgers University in New Jersey, says, in his book The Atoms of Language that there are similar atom-like constructs that combine in various ways to produce the 6,000 languages spoken in the world today. He believes that right now, at the start of the 21st century, in terms of unravelling the atoms underlying language we are at the same place where 19th century chemists found themselves, with just a basic understanding of the first few elements. The epiphany for chemistry came when the Russian, Mendeleyev, enunciated the Periodic Table of Elements, which not only lined up all the known elements of the day but also helped predict what other elements would be discovered subsequently.

How does it help us to unravel the "atoms" that make up the world's languages? The argument goes something like this: in advanced economies virtually all economic growth today comes from the service sector. Competitive advantage in the service sector, particularly in "producer services", will lie in improving productivity, the output of each service industry worker. This can be achieved only by automating some parts of service delivery. Take for example, a call centre, where an operator receives a phone call from a customer, say, enquiring about a delivery date. The operator checks in a computer database for this information and orally gives the customer the delivery status. Essentially, the process involved a voice call that has been "understood" by a human operator who then keys in a query to a computer data base. The answer returned by the computer is then visually read by the operator and relayed orally back to the customer. Now imagine a device that records the customer's voice query , converts the voice request into text, feeds the text into a search engine, which then searches the database for the answer, and, having found the answer in text form, converts it back into a voice message to be read back automatically to the customer. This device would dramatically improve the productivity of the call centre. All that stops us today from doing this today is that there are far too many errors in one step""that of converting the incoming voice message into text.

Unravelling the underlying "atoms" of languages would greatly help in improving the accuracy of these voice-to-text conversions.

The application of this knowledge would extend far beyond customer service. Other applications include machine translations of books (imagine the benefit of automatically translating books from English into all the Indian languages), and monitoring telephone conversations for terrorist activity. Virtually all aspects of the service economy involve hearing human language, converting this to text, using key words from this text to search a database and returning back some information back in voice form.

Whoever unravels the atoms of language first would have a great competitive advantage in building mega business around this knowledge. Should we not be spending as much effort on unravelling the atoms of language as we are spending on electric power from the earlier atom?

Comments welcome at ajitb@rediffmail.com