Simon and Phil, (01)
This thread touches on major controversies in AI, NLP, psychology,
neuroscience, and cognitive science in general. (02)
Simon
> That is not to say that the hypotheses are wrong- just that
> [Phil's] example is not strong evidence. (03)
Phil
> I didn't claim this example [copy below] was proof or strong
> evidence of any particular model of language acquisition.
> It's just an example of the importance of pragmatics and
> semantics for child language acquisition. (04)
It's rare that any single example can refute any well developed
theory. But that example illustrates critical aspects of language
that Chomsky considered outside the scope of "scientific" linguistics:
pragmatics (AKA performance), background knowledge derived from non-
linguistic sources, and the continuity between language learning
and other methods of learning. (05)
In the 1950s, Chomsky drew a sharp distinction between competence
and performance in order to carve out a niche for formal linguistics.
He compared his strategy to the practice by physicists, who studied
simple examples in order to formulate their theories. (06)
Chomsky also made the assumption that humans have a genetically
inherited LAD (Language Acquisition Device) that is based on some
UG (Universal Grammar) that is unique to humans. (07)
Many linguists (including a few computational linguists) liked those
two hypotheses because they enabled them to ignore the messy data
and formulate clean, elegant theories. Unfortunately, those theories
weren't very useful for interpreting human performance or for
implementing computer programs that could. (08)
Meanwhile, psychologists never liked Chomsky's methods because the
messy data he ignored was their primary subject matter. Today, many
linguists have come to agree with the psychologists: the messy data
holds the key to the way language works. (09)
Chomsky developed various theories over they years, but they're
all designed to support his two basic hypotheses. For years, many
linguists took them seriously, but their numbers have been dwindling. (010)
Researchers in other branches of cognitive science considered them
a step in the wrong direction. They created artificial barriers
between the methods of linguistics and the methods that have proved
to be useful in the other fields: psychology, philosophy, AI,
neuroscience, and anthropology. (011)
Even computational linguistics ignored that distinction -- because
the data they process comes from performance, not competence. (012)
Summary: I believe that Chomsky's work and the competing methods
by formal linguists such as Montague, Kamp, and others have made
important contributions to the field. But I don't believe that
they are adequate to explain the human capacity for learning,
understanding, and using language. (013)
I follow Peirce's principle: "Do not block the way of inquiry."
If somebody has a strong intuition based on the methods by Chomsky
or Montague, I'd encourage them to pursue it. But I would *not*
consider their methods more worthy or "scientific" than many
other methods developed in cognitive science. (014)
John
___________________________________________________________________ (015)
Example by Phil Jackson: (016)
In a talk about child language acquisition, Paul Vogt of Tilburg
University first plays a soundtrack of a woman saying something in a
foreign language, and asks the audience to guess what she's saying. (017)
No one can guess, since the language is from Africa, and the audience
speaks only European languages. He then plays a video, which shows the
woman standing in a small village with a child perhaps only 3 years old.
She points to a cup on the ground and points to a bucket of water, and
says the words no one in the audience could understand. It's now clear
from the visual information that what she is saying means "Give me some
water". The young child picks up the cup, fills it with water from the
bucket, and gives it to her mother. (018)
This simple example illustrates the relative importance of pragmatics,
semantics, and syntax, for child language acquisition. The pragmatics
(context information including visual information about the scene and
gestures) enables understanding the semantics of speech, when a child is
learning a language and doesn't yet understand syntax. (019)
Any infant with normal intelligence can learn any natural language, if
it is raised by a parent who speaks that language. Yet adults cannot
directly describe the syntax of a natural language to an infant, and
there is a very wide space of possible language syntaxes. So very young
children need to use pragmatic information to learn the semantics of
speech in a language, and to also gradually learn the syntax. Anyone
fortunate enough to watch a child learn to talk can observe that
understanding pragmatics and semantics precedes and enables acquisition
of syntax, for very young children. (020)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (021)
|