First of all, let me say that I am very envious of both Jasper and
Nik for being able to attend and participate in the construction
grammar conferences (or workshop, as the case may be).

Here are some rather jumbled comments on Jasp's paper.

Jasper Holmes wrote:
>
> Figure 1: WG isa: DG, cg, CG
>
> a. WG isa Dependency Grammar (DG):-
> words (and the relationships between them) are the only primitives of
> linguistic structure

In Fillmore & Kay's construction grammar (CG), this is (normally)
expressed expressed by the [val { }] feature. What corresponds to
word strings in WG, CG expresses as [lex -] (whereas single words
are labeled [lex +]). The main difference between CG and WG is
that WG really has nothing that corresponds to [max +] or [max -]
(i.e., phrasal and nonphrasal constituent, respectively).


> b. WG isa cognitive grammar (cg):-
> (knowledge of) language is a subpart of the conceptual structure that
> constitutes more general knowledge ('the mind')
> [this means principally that we don't rule out, a priori, any
> information from participating in linguistic structures, and cuts both
> ways: not only do we see it as our responsibility to account for all the
> properties of linguistic structures (cf. Borsley abstract), but, on the
> plus side, we allow ourselves to use any appropriate information (eg
> 'encyclopedic' semantic properties) to help explain linguistic
> phenomena]

Also: WG recognizes the usage-based nature of grammar (e.g., frequency
effects), as well as prototype-based categories (e.g., default
inheritance). Furthermore, WG organizes grammar in terms of interlocking
networks (see Jasp's point 1.2).


> c. WG isa Construction Grammar (CG)
> all regularities of language are represented in a constructicon
> (Jurafsky 1992), a network of more or less schematic lexical
> representations (Fillmore 19??, ix: "an approach to language according
> to which there is no sharp distinction between the lexicon and the
> non-lexical parts of grammar.")
> [this is a claim and means we have to find a plausible account, for any
> pattern we find, in terms of an individual class of words or
> dependencies (see (a)); (for advanced students:) the difference between
> lexical structures based on word( classe)s and those based on
> dependencies/y classes is not so great as you might think, it's really
> just a matter of perspective: an object must isa Noun/some Nouns are
> objects]

As you say in your abstract, "the essential difference between lexical
items and grammar rules is simply one of generality [...]." I would
further add that the difference of generality is not categorical, but
one of degree. This becomes clear when you consider conventional
expressions.


>         a. Lexical properties of DOG: spelled /dog/, refers to dogs (of both
> kinds), isa Noun, ...

I know that this is nitpicking, but shouldn't it be "spelled <dog>."

Hope this helps,

Joe
__________________________________________________________
Home page: http://lingua.fil.ub.es/~hilferty/homepage.html