Re: Tiny Grammar
| From: | Edward Heil <edheil@...> |
|---|
| Date: | Tuesday, March 16, 1999, 16:05 |
|---|
Well, since nobody has ever succeeded in fully specifying the grammar of
any natural language, it's not exactly an unspeakable crime that what
somebody considers a reasonably full specification of a grammar actually
leaves a few things out.
Personally, I suspect that there is in the real world only one rule of
grammar: integrated expressions (phonology) correspond to integrated
content (semantics). Everything else arises from the nature of the
units of phonology and semantics. Regularities in grammar in natural
languages arise from the fact that it is easiest to use and remember
inventories of constructions which are systematically related to each
other, rather than lists of utterly unrelated constructions.
("Construction" here is a superset of "word" and refers to any link
between expression and content.)
Charles Fillmore, Ronald Langacker, George Lakoff, and others have
pursued this kind of scheme with great fruitfulness in analyzing natural
languages.
Under this scheme, there is no strict dividing line between the grammar
and the lexicon; the grammar lives in and is intertwined with the
lexicon. This is not to say that all the grammar is tied to specific
words; *patterns* of word-types are considered to have their own
sematics as well. The classic example of this is the caused-motion
construction analyzed by Adele Goldberg in _Constructions_:
NP V NP PP
ex.: "he sneezed the napkin off the table" -- where the construction
itself is shown to have a definite semantic value which integrates with
the semantics of the words involved.
I'm babbling now. Sorry. :)
Gary, Rubin's interesting. I don't have a problem with the grammar
being short, because I can imagine the vast majority of the work being
done by the "inventory of constructions" or lexicon. If anything, the
grammar seems a bit restrictive. But then, I suppose it's supposed to
be, as it's supposed to make everything machine parsable.
I like it a bit better than Allnoun in that it doesn't have those silly
parentheses, but I see the fact that the slots are all predefined as
perhaps a bit of a limitation.
Ed
>From: Nik Taylor <fortytwo@...>
>Gary Shannon wrote:
>> Can the complete comprehensive grammar textbook of a conlang be
written on
>> the back of a single postcard?
>
>No. Your rules need much explanation, for one thing. Also, the idea
of
>each word having a unique pattern is somewhat suspicious. In reality,
>it would be more reasonable for parts of speach (which, as you say, is
>not actually eliminated) to have common patterns, which would be
>included in the grammar. In addition, there are very likely a number
of
>unstated assumptions you have, which may not be obvious to another
>person.
>
>--
>"It's bad manners to talk about ropes in the house of a man whose
father
Get Your Private, Free Email at http://www.hotmail.com