Revision 1 as of 2014-02-18 04:48:34
|Deletions are marked like this.||Additions are marked like this.|
|Line 1:||Line 1:|
|#acl WoodleyPackard:read,write,admin DanFlickinger:read,write,admin EmilyBender:read,write,admin FrancisBond:read,write,admin StephanOepen:read,write,admin AnnCopestake:read,write,admin||#acl WoodleyPackard:read,write,admin DanFlickinger:read,write,admin EmilyBender:read,write,admin FrancisBond:read,write,admin StephanOepen:read,write,admin AnnCopestake:read,write,admin All:read|
This discussion took place over a coffee break, so no live note taking.
Context: discussing the possibility of having the pronoun referential indices for the possessor and the main subject be identical in an idiom like "miss one's mark" or "look as though butter wouldn't melt in one's mouth" (we suspend disbelief about that one being part of our language for now)
1. various attempts to come up with an example where doing so would run afoul of scoping rules.
something like: "All children can't be trusted, even looking as though butter wouldn't melt in their mouths."
If the subject of _look_v_rel were "children", then the its quantifier must be high -- (bad why?)
2. what about getting the info out of syntax about what can and can't corefer? Dan's allergic to binding theory, not much hope of those constraints emerging.
3. some discussion of what a delph-in-sanctioned representation of coreference links should look like, that didn't reach much of a conclusion
Some further recollections from Emily:
The reason we were worrying about those crazy examples was to try to validate or obviate the concern that having one quantifier bind variables so far apart in the MRS might lead to structures that were untenable in some way.
One question was whether the language for representing coreference groups (chains etc) should be the same as that used by the grammar for representing constraints on possible/impossible antecedents given (what we can model of) binding theory. I think the consensus was no, they didn't have to be the same, but the discussion of what the coreference group/chain representation should look like was inconclusive. It turned partly on split-antecedent examples: Kim likes Sandy and they went on a date tonight. The question here is whether the it is enough to say that they denotes a group and assert that the referents of Kim and Sandy are both members of that group, or if we should construct the group separately and then have it be the antecedent of they, making the treatment of the they instances in examples like the following more consistent: Kim likes Sandy. They went on a date tonight by the end they had already made plans for more dates and they both seemed happy.