Robotics Reference
In-Depth Information
-7-
How Computers Communicate
Natural Language Processing
Natural Language Processing (NLP
1
) is the branch of Artificial Intelli-
gence concerned with enabling computers to talk like you and me, to
understand what is said to them, to be able to conduct sensible conver-
sations and even to translate into and out of foreign languages. When
computers can understand what we mean when we speak or type some-
thing in English or in any other natural language, they will be much
easier to use and will fit in more with our everyday lives. That is why,
ever since Alan Turing first described, in 1950, what is now known as the
Turing Test, this challenge has been widely regarded as the touchstone of
AI. But the goal of having computers engage in intelligent conversation
appears to be almost as elusive now as it was then.
Some of the Problems in NLP
Why is this task so hard? After all, our children can make more sense in
their conversation at the age of three or four than can the biggest, most
powerful computers of today, even when running software that is the
product of tens of thousands of person-years of research or more. The
resources that have been applied to NLP exceed many times over the
resources that have been applied to computer Chess, and this includes
some of the brightest minds on our planet, not only from the field of
computer science but also from linguistics, mathematics, statistics, psy-
chology and other areas of cognitive science. But despite all this effort,
while a program can defeat the world's strongest Chess player, no pro-
gram can conduct even a half-hour long conversation at the level of a
high school freshman. Why? The word processing software on our per-
sonal computers often appears to be smart enough to correct our poor
use of language, putting a green wavy line on the screen to warn us when
1
In this volume the abbreviation NLP is used only in this sense, not to be confused with the
more recently born and totally different discipline of Neuro-Linguistic Programming, for which the
same abbreviation is commonly employed.
