Dick:
> Jasper:
> >> DRINK has an object. Sometimes it doesn't, and that's not a
> >> particular problem either, but *sometimes* when it doesn't have an
> >> object it means specifically 'drink alcohol': _John doesn't drink_. Now
> >> in order to be able to represent this lexically (I think) we
> have to say
> >> that there's a special subcase of DRINK, which means 'drink
> alcohol' and
> >> which MUST NOT have an object.
> >>
> >> So this isn't a case of the more specific object of the token
> >> overriding the general specification associated with the type, but of
> >> the more specific category banning a property of the general category
> >> outright.
> >>
> >> Can't quite pin down exactly why this makes me uncomfortable,
>
> ## I can't see why it should make you uncomfortable at all. Why can't a
> default `X has a Y' be overridden? (e.g. DRINK has an object - but
> DRINK/alc has no object) Even if it's a problem in HPSG it doesn't seem to
> be one in WG.
I agree. And I don't know of any cases where indefeasibility is required.
--And.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|