[matrix] GTOP in Matrix, was: Re: Message-free Matrix

[Emily M. Bender]

For the "better late than never files..."

We have finally gotten around to fixing this, and the fix (following
Francis's suggestion) is now in the live Matrix site.  (Thanks to Joshua
Crowgey for the actual implementation there.)

And of course we were able to reproduce the behavior that
Francis notes below.  The feature structure has the LTOP
that we would expect (linked to the main verb's LBL, or to
a scopal modifier, if there is one).  In the MRS displays, however,
this is replaced with the GTOP.

I take it from Ann's reply (pasted below Francis's message
below) that this is in fact the expected behavior, but I wanted
to double check on this.  Are there cases where we would
prefer to see the LTOP (rather than GTOP) in the various
MRS views?  If so, is there something we should put in
e.g., user-fns.lsp to make it so?

Emily

On Mon, Dec 31, 2007 at 6:20 AM, Francis Bond <fcbond at gmail.com> wrote:
> G'day,
>
> I made a grammar with the new matrix, and the MRSs don't scope.   If
> you explicitly add a global top (GTOP), separate from the local top
> (LTOP), they then scope once more [thanks to Dan Flickinger for this
> solution].
>
> ;;;Changed to (in mrsglobals):
> (defparameter *psoa-top-h-path*
>  `(,(vsym "HOOK") ,(vsym "GTOP"))
>  "path to get the hook's ltop from an mrs fs (psoa - parameterised
> state of affairs -
>  old terminology)")
>
> ;;;(and added GTOP to the MRS definition in matrix.tdl)
> hook := avm &
>  [ GTOP handle,
>   LTOP handle,
>   INDEX individual,
>   XARG individual ].
>
> Unfortunately, this means that the hook of the MRS is the GTOP, and it
> is not explicitly linked into the MRS at all, which makes much of the
> MRS manipulation we did when extracting hypernyms from definitions
> slightly more complicated.  If anyone knows of a good way of linking
> "_neru_v_rel"the GTOP and what used to be the Hook's LTOP, please let
> me know.
>
> In more concrete terms, the thing called the LTOP in the matrix below
> (which I think is really the GTOP), is not now linked to anything
> inside the MRS.   I had expected it to have some relation (h1 qeq h8)
> to the LBL of "_neru_v_rel", whose index it shares.
>
> inu ga neru
>  [ LTOP: h1
>   INDEX: e2 [ e SF: PROP-OR-QUES E.TENSE: TENSE E.ASPECT: ASPECT
> E.MOOD: MOOD SORT: SEMSORT ]
>   RELS: <
>          [ "_inu_n_rel"
>            LBL: h3
>            ARG0: x4 [ x SORT: SEMSORT PNG: PNG COG-ST: COG-ST SPECI: BOOL ] ]
>          [ "unspec_q_rel"
>            LBL: h5
>            ARG0: x4
>            RSTR: h6
>            BODY: h7 ]
>          [ "_neru_v_rel"
>            LBL: h8
>            ARG0: e2
>            ARG1: x4 ] >
>   HCONS: < h6 qeq h3 > ]
>
> Yours,
>
> --
> Francis Bond <http://www2.nict.go.jp/x/x161/en/member/bond/>
> NICT Computational Linguistics Group
>

On Wed, Jan 2, 2008 at 12:25 PM, Ann Copestake
<Ann.Copestake at cl.cam.ac.uk> wrote:
>
> If what you want is really the LTOP, then the option I would suggest
> is to use the mrs algebra code which gives you the hook structure as
> well.
>
> Ann
>
>

-- 
Emily M. Bender
Associate Professor
Department of Linguistics
Check out CLMA on facebook! http://www.facebook.com/uwclma