[matrix] Pronoun Relations

[Ann Copestake]

Antonio.Branco at di.fc.ul.pt said:
> But it is worth keeping in mind that the expensive procedure of anaphora
> resolution (with sub-optimal accuracy given current methods) may not be
> needed in most cases. Instead of trying to determine what is the antecedent
> of the anaphoric expression at stake (phonetically overt or not), we may only
> need to provide a proper characterization of its anaphoric capacity (i.e. of
> the set of its admissible antecedents, i.e its anaphoric type: short-distance
> reflexive, pronoun, etc): in many cases just translating it into an anaphor
> of the target language with identical anaphoric capacity will do the job.

I completely agree with this - we want to be able to leave distinctions 
underspecified.   But for some of the projects we're involved in, we need to 
be able to resolve and having to do this by going back to the parse tree would 
be very awkward, especially in the RMRS approach, where we're dealing with 
multiple different sources of the RMRS.