Monday 15 March 2010

On Computational linguistics

King’s College, London is proposing to get rid of, among others, three people who do computational linguistics after deciding that they are surplus to requirements. Their methods have been a little shabby and coercive, their thinking has been muddled. I attempt here to explain why this is a remarkably dumb decision.

There is no denying that computational linguistics is a slightly weird discipline which is difficult to pigeon-hole.

The people who work in the field frequently speak and think geek and are not given to explaining themselves in ways others can easily understand. Many are irascible, impatient and obsessive. They are frequently difficult to integrate into a well-oiled and controlled organisation of the sort beloved by modern professional managers. (I speak generally here; I have not talked to any of the people at King’s, only to colleagues and associates abroad).

Of the three, two (Shalom Lappin and Wilfried Meyer-Viol) are in the King’s philosophy department, one (Jonathan Ginzburg) in a Computer Science now turned into Informatics department, which makes it easy to target them as useless.

Such a conclusion, however, misses the way such undertakings are dissolving the traditional barriers not only between disciplines, but also between the arts and the sciences. The result is one of the most exciting and innovative undertakings going on in academia today, something to be encouraged rather than eliminated.

The influence of such work spreads far – which is one of the reasons protests about their removal has aroused such wide condemnation. Theirs is one of the key areas of research that will fuel not only the next generation of advance in computer technology, but the generation after that as well.

This will be artificial intelligence, which after many years of promising wonderful things but not delivering much, is finally beginning to turn into a field which is making real advances.

To do this has required drafting in people from many different disciplines. To teach a computer how to tell the difference between reality and a reflection, for example, can and does call on the insights of software designers, engineers, neurologists, cognitive psychologists, specialists in robotics, linguistics and philosophy. The spin-off is not only smarter machines, but also a better understanding of brain function.

These collaborations are often informal, unfunded and (crucially given the state of British university management) happen without the knowledge of administrators who like to count these things up so they can reward people for “impact.”

The same goes for verbal understanding. Teaching a computer not only to analyse and understand words, but grasp meaning of speech and writing, the implicit and explicit texts and sub-texts to human utterance, cannot be done by someone who simply writes software. It needs input from people who understand the structure of language and how the human mind interprets those structures. This is why computational linguistics is of central importance.

Out of it, you get three main lines of advance: firstly in the field of human cognition, how we understand and process language, which has application for the treatment of brain malfunction. Secondly, in pure linguistics itself, the more philosophical understanding of how language works. And finally in computer technology. In the last field the goal is instant computerised transliteration, accurate translation between languages, and machines that can understand and respond in ever more subtle and sophisticated ways, rather than being glorified speak-your-weight machines. Machines that can read, speak and, most importantly of all, understand.

That’s their impact, even if they are only in a philosophy department. Computational linguistics is at the heart of research with the potential to become one of the great – and most profitable – technological leaps forward of the next few decades.

King’s has some of the acknowledged experts in the field. An organisation which thought in the long-term would take away their passports, lock them in a room to get on with their work undisturbed and, if they were threatened, beg in the streets to get the money to keep them.

Alas, it thinks differently. A strange letter which appears to have been sent from one King’s administrator to another, and which went astray, sums up the problem: “[Ginzburg’s] work is at the periphery of computer science.”

This misunderstanding typifies an attitude which has meant that, while the United States can produce companies like Google, which harness advanced mathematics to dominate the world (and make a lot of money), Britain’s most famous innovation in the last couple of decades has been the bagless vacuum cleaner.

As Talleyrand put it, this is worse than wrong. It is a blunder.

-- Iain Pears

1 comment:

  1. Thank you for an eloquent defence of this subject. I've just looked at Jonathan Ginzburg's webpage. To consider someone taking a computational approach to dialogue, and producing work of Ginzburg's standards, to be "at the periphery of computer science" is gobsmackingly laughable.

    ReplyDelete