from empirical side to recipe for disaster

[Detmar Meurers]

As requested in the PS I'm forwarding Mark's reply on the ANY thread:

----- Forwarded message from Mark Johnson <Mark_Johnson at Brown.edu> -----

Hi Steve and Detmar,

It's good to hear from all of you.  Detmar is of course correct when he says
that we can't expect other people to clarify the meaning of our theories, and
that we need to guard against accidentally saying something inconsistent.  I
personally think that a formal semantics is a great way to accomplish these
goals.

I wish I could give a simple declarative semantics for ANY values.  I suspect
that if one could work out a clean declarative semantics for ANY values, we
could probably simplify and generalize the basic idea so that it integrated
better with treatments of default values, negation and the type inheritance
hierarchy.

Regarding LFG, there are a number of things in the LFG formalism that I
personally would like to see clarified.  Inside-out functional uncertainty is
something I would like to see a clean formal semantics for, for example.

Would LFG be a better theory if it had a clean, simple declarative
semantics?  Of course yes.

Should we ignore LFG because it doesn't have a fully worked out declarative
semantics?  Not at all.

Is working out a formal semantics an important thing to do?  Until it is
done, I don't see how anyone can know the the answer to this.  It could turn
out to be incredibly important; e.g., a formal semantics might lead to
generalizations of the current LFG mechanisms that turn out to be
linguistically very useful, or it could turn out (horrors!) that current LFG
is inconsistent.   I personally think that the meaning of ANY values is
sufficiently clear that working out the details is unlikely to yield dramatic
surprises, and that there are more productive things to investigate (like how
one can define probability distributions over LFG or HPSG linguistic
analyses, so you can use statistical tools for learning and modelling
preferences).  But I would be happy to be proven wrong (smile).

All the best,

Mark

PS.  I guess I really ought to ping the HPSG list people to get put back on.
Thanks for forwarding my messages there!

----- End forwarded message -----