[Corpora-List] Is a complete grammar possible (beyond thecorpus itself)?

[John F. Sowa]

Mike and Rob,

MM> I meant to ask... whether in these two pieces that you
 > quoted from, Kamp and Partee aren't *now* saying something
 > that is quite compatible with Chomsky's claim of the autonomy
 > of syntax.  Specifically, aren't their comments that formal
 > semantics has not worked out (at least that's what I take them
 > to be saying) compatible with Chomsky's view that syntax is
 > an autonomous module, and formal?

I seriously doubt that anyone, even Chomsky, still believes the
original claim:  A formal grammar generates *all and only* the
sentences of an NL that native speakers consider acceptable.
Even Culicover and Jackendoff (ref's below) now claim that more
collaboration among different subfields is important.  If they
believed in the autonomy of syntax, they wouldn't say that.

RF> We can see how the language gradually generalizes, but in
 > a context specific way.

I agree, and we may be talking about similar mechanisms, but with
different terminology.  I think we agree on the following points:

  1. When a system (human or computer) finds an unrecognizable
     grammatical pattern, it should make its best effort to find
     some interpretation for it.

  2. But it should also keep some kind of record of the original
     pattern and the interpretation made.  That is necessary
     in order to make a generalization, if the same or similar
     pattern occurs again.

  3. If another unrecognizable pattern comes along, the system
     should check whether there were earlier patterns like it
     in whatever storage is used for "nonce grammar" instances.

  4. As more examples of a nonce-grammar pattern accumulate,
     its status increases from "probable error" to "temporary
     innovation" to "common in the genre" to "standard".

  5. The above points imply that some kind of storage is
     required for every unrecognized pattern -- at least
     until it has been assimilated into some encoding that
     is similar in nature to the encoding of whatever is
     typically called the "standard" grammar.

If this is what you mean, then I think we are in agreement
about some principles of language that are quite different
from the so-called "mainstream" of 20-th century linguistics.

RF> There is a vague idea we have to merge lexical and syntactic
 > aspects of text, but no one has a clue how to do that.

I would say that there are many clues, many proposals for doing
different kinds of mergers, but not enough evidence to make a
good recommendation about which one(s) to choose.

I am encouraged by a steady stream of recent publications
that indicate the "mainstream" is creeping along in this
direction.  Following are a few (check Google for full ref's):

  - _Simpler Syntax_ by Culicover & Jackendoff (2005) is a
    recognition by long-time Chomskyans that a major overhaul
    is long overdue.  However, they are still trying to preserve
    a very large part of the results obtained by the Chomskyan
    linguists in a way that is fairly conservative.

  - _Dynamic Syntax_ by Kempson, Myer-Viol, & Gabbay (2001)
    is a more radical approach to syntax, but the semantic
    theory by Gabbay is a very formal logic-based approach.
    Gabbay uses "decorated trees" instead of a linear notation
    for the logic, which I like, since conceptual graphs can
    be viewed as "decorated trees" glued together in similar ways.
    But I believe the logic should be as dynamic as the syntax.

  - _Cognitive Linguistics_ by Croft & Cruse (2004) combines
    radical construction grammar (RCG) with lexical semantics
    in a way that makes both more dynamic than the above approaches.
    RCG does allow syntax to evolve from more primitive patterns,
    but Croft and Cruse don't say how it would be possible for
    logic to evolve.

The fact that children by the age of 3 use words for the logical
operators (e.g., 'and', 'not', 'some', and others) indicates that
logic somehow evolves out of the infant's early one and two-word
phrases.  And the fact that all mathematicians, logicians, and
computer programmers use NLs to explain the most abstruse theories
imaginable indicates that there is no limit to how far the
expressive power can evolve.

John



_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora