Filler-gap mismatches

[Berthold Crysmann]

Carl Pollard wrote:

> Hi Ron,
>
> Now I owe you TWO long answers. But again, for now just a brief
> remark. When I suggested moving the category into the f-structure,
> what I really meant was: first divide the category into part-of-speech
> and projection-level (or bar-level if you prefer), and just move the
> part-of-speech part into f-structure. That way, all the nodes along the
> head chain can have the same f-structure, as you want. Then f-structure
> would correspond pretty closely to the combination of HEAD and ARG-STRUC
> which according to the correct splinter group of HPSG is the stuff that
> is shared up head paths.

Dear Carl,

Do you have a story about the composition of ARG-ST values on the mother
node of a coordination structures? (That was one of the nastier arguments
that cropped up in the lengthy thread on this topic a few weeks ago).  If
so, I'd be very happy to know what your solution looks like. If one wants
to assimilate HPSG's ARG-ST to LFG f-structures, the fact that the former
but not the latter is an ordered list may be a major obstacle, no?

Berthold


> Another splinter group, which I will
> neutrally characterize here as the `incorrrect' one, has it that only
> HEAD is shared up head paths, while ARG-STRUC appears only on lexical
> heads.
>
> Carl

--
Berthold Crysmann
Deutsches Forschungszentrum Kuenstliche Intelligenz (DFKI) GmbH
Stuhlsatzenhausweg 3
D-66123 Saarbrücken