[developers] portuguese hpsg

Berthold Crysmann crysmann at ifk.uni-bonn.de
Tue Apr 14 11:13:49 CEST 2009

Dear David Raposo, 

this is a known issue with MRS and symmetric predicates. 
Within MRS proper, there does not seem to be any straightforward
solution. Both the ERG and the GG make a somewhat arbitrary canonical
decision (mapping the subject to ARG1). For German, I often get two
readings with identificational copulae. 

What I'd suggest is to use sense tagging to identify predicates that
display this behaviour so they can be detected downstream. 

E.g. _marry_v_sym_rel or _be_v_sym_rel (as in "The winner is ...)

You can then write MRS transfer rules to relate the different uses. 
There is support for regular expressions on PRED values in the transfer

I am cc'ing your mail to the DELPH-IN developers list. 
Hope this answers your question.



On Tue, 2009-04-14 at 09:47 +0100, David Raposo wrote:
> Prof. Berthold Crysmann,
> Hi, my name is David Raposo and I'm working on a HPSG grammar for
> Portuguese with Prof. António Branco, who gave me your contact. He said
> you might be able to help me in the problem I'm facing with right now.
> I'm trying to find a good semantic representation for verbs like
> "marry", which have a reciprocal behavior.
> We may say
>    A married B
> and use the following logical form
>    marry(A,B)
> but this way we are losing the information that B married A too. This
> way we won't be able to represent sentences like
>    A and B got married (or They got married)
> which have the same meaning of the first one. In portuguese, and I think
> in English too, this last one also might also mean "A married x and B
> married y".
> What do you think it would be a good representation for this kind of
> behavior?
> Thank you and best wishes,
> David Raposo

More information about the developers mailing list