Default unification: online vs. offline

[Rob Malouf]

Luis,

I don't think there are serious computational or formal problems with using
persistent defaults.  The LKB implementation of Lascarides and Copestake's
(1999) YADU operator is quite fast and robust, and for now at least it looks
like if unification is fast enough for your purposes, then default
unification probably will be fast enough too.

Whether you *ought* to use persistent defaults is another question. I claim
in my dissertation that the Cognitive Grammar notion of categorization is
best cast in HPSG in terms of inheritance hierarchies, using something like
Koenig's on-line type construction.  On the other hand, Ann Copestake has
argued that the interface between pragmatics and lexical semantics involves
the interaction of persistent defaults. So, if you've got good evidence that
persistent defaults really provide the best solution to some problem, you
shouldn't let computational fears prevent you from working out the details or
even attempting an implemention.

------------------------------------------------------------------------------
Rob Malouf                       Alfa-Informatica, Rijksuniversiteit Groningen
malouf at let.rug.nl                                http://www.let.rug.nl/~malouf