Increasing interest in the HPSG conference

[Carl Pollard]

Hi everyone,

I think Paola's suggestion to merge the FG and HPSG conferences in
2005 makes a lot of sense. I hope at the Leuven meeting consideration
will be given to following that model again.

When that was done in 1998, CG was also a component. It would be good
to hear from some CG-oriented subscribers to this list about whether
there is interest in the CG community in joining forces in this way
again.

Some people have suggested that HPSG should team up with other
nonderivational frameworks, in order to form a united front against
the common enemy. But derivationality is not the enemy, pace Pullum
and Scholz. CG is also derivational, and so is some recent formally
explicit work inspired by the MP that deserves a hearing in the
nonderivational community. Enemies are authoritarianism,
narrow-mindedness, vagueness, to name a few.  I suggest it is more
constructive to make precision/explicitness the key criterion, not
nonderivationality. (For the old-timers among you, recall the first
couple of WCCFLs as embodiments of the spirit of free but careful
inquiry.)

It's also been suggested that the line be drawn to favor restrictive
formalisms at the expense of expressive ones, and that HPSGers should
correspondingly get back in touch with their their GPSG roots. I agree
that it's illuminating to reconsider now the architectural choices
that were made 20 years ago that distanced HPSG from GPSG (and
CG/Montague grammar), but not because GPSG was a constrained
framework: it wasn't, as Uszkoreit and Peters proved. (The context-
freeness of GKPS was bought at the cost of imposing a linguistically
unmotivated prohibition on application of a given metarule more than
once in a derivation.)  Time after time in the history of our field,
seemingly (or even truly) spartan formalisms have turned out to
generate all r.e. languages off-the-shelf (or once they are enriched
with mechanisms to handle complex phenomena gracefully). The reality
is that (as most recently discussed by Jim Rogers) both restricted AND
expressive formalisms are needed, for different things.

More specifically (in connection with HPSG vs. GPSG) it has been
suggested that moving from trees to feature structures was a
misstep. Maybe so, but it is a short step, if trees can have their
nodes labelled by AVM's and some of the features of those AVM's have
values (such as indices) that can show up at more than one node. A
more important distinction than whether the tree is gussied up into a
feature structure is whether the tree is viewed as (a) a model
(structural representation) of a sign, or (b) as a proof (derivation)
of a sign. I believe this is a (maybe even THE) fundamental issue that
has to be clarified before the competing frameworks can become
scientific allies (not just organizational ones aimed at economizing
on business travel and sharing conference facilities).

One last point about formalisms. There's been much talk about the
success of the MP in terms of number of practicioners, as compared
with competing frameworks. Of course it doesn't hurt to be led by the
one and only bona fide celebrity linguist. But there is another
extremely important ingredient in the popularity of the MP and its
predecessors that has not been mentioned: unless you are trying to
formalize it, it requires ABSOLUTELY NO TECHNICAL KNOWLEDGE OF ANY
KIND: no algbera, no proof theory, no model theory, no formal
languages, no automata, no algorithms, no data structures, nothing
except maybe earlier versions of the MP. That means that there is a
certain kind of audience/consumer/prospective recruit for whom the
more formally explicit frameworks cannot hope to compete. At the
opposite end of the spectrum is CG, which really does require some
technical knowledge on the part of the practicioner, but at least the
technicalities involved are well-explicated, widely-known, and
long-established ones; so it is easy for the technically inclined to
be enticed into it.  HPSG occupies the uncomfortable middle ground:
it's hard to do it right without understanding the underlying
formalism, but the underlying formalism is home-brewed, of recent
origin, hard to understand, and disconnected from mainstream ideas in
math, logic, and computer science.  So it SCARES off the technically
uninclined and TURNS off the technically sophisticated. Or, to put it
another way, it is as hard as CG but as intellectually isolated from
adjacent disciplines as the MP.

Carl