[matrix] prpstn_m_rel and mrs2rmrs
Anette Frank
frank at dfki.de
Thu Sep 8 11:13:41 CEST 2005
Hi Dan, Ann (and others),
following Dan's announcement at the Delph-In meeting, that the standard for
coding of messages will soon be (or did already?) change, we would like to
make this adjustment also to our German RMRS bank before releasing it.
Before doing this, though, we would like to clarify the following issues:
- To be sure: we assume that for the RMRS, the new representation
should look
like this: (Although I *am* on the Matrix mailing list, I haven't
seen any
discussion of this - nor is there one obviously present in the Archive)
prpstn_m_rel
LBL h1
ARG0 e1 (index/variable of the verb)
MARG h2
- At present, the conversion from (HPSG-produced) MRS to RMRS falsely
converts
the (current) MARG of message relations to ARG0. That is, for the new
representation,
we will have to correct these conversion scripts, to avoid
confusion/conflicts.
As far as I can see, MARG should be allowed as a name for rargs in the
RMRS DTD,
so there shouldn't be a problem having MARG in the RMRS.
So, can we safely assume that these steps will be taken, and change our
structures to
the new standard?
Regards,
Anette and Kathrin
From Delphin-Wiki
> Some Matrix changes are in response to requests, and some harmonizing
> still takes place to existing Matrix stuff:
>
> *
>
> Predicate names for messages (see Matrix mailing list for
> discussion)
>
> *
>
> Messages will have an ARG0 that will be linked to the INDEX
> event of the clause that supplied the message.
>
> *
>
> (Messages are quite a complex system that does very little work
> -- it's a three-way distinction at the moment.)
>
More information about the matrix
mailing list