Mike Calcagno (calcagno@sfs.nphil.uni-tuebingen.de)
Tue, 6 Aug 1996 12:56:46 +0200 (MET DST)
hi folks. > If we agree that the principles are stated as "implicative > constraints within a certain feature logic" (in other words, the > formalism is "based on typed feature logic with recursive types and > relational constraints"), then each principle has the form: > if X is compatible with description_1 > then X must be compatible with description_2 > We must clarify what "compatible" means and what the description > language is, but I want to point out a specific problem. just a quick note on the above point: one nice thing about King's SRL is that negation is classical... so, in the denotational semantics for formulas that he provides, D(f_1 --> f_2) = D(~f_1) U D(f_2) = U\D(f_1) U D(f_2) where U is the universe of objects, and \ is set complement (oh, i also used U for ascii-fied union, but i think the context makes the difference clear)... > Consider the subcategorization principle of P&S94: > if X is a headed phrase, > then DTRS|HEAD-DTR|SYNSEM|LOC|CAT|SUBCAT is > the concatenation of SYNSEM|LOC|CAT|SUBCAT with > the list of SYNSEM values of DTRS|COMP-DTRS > Can the consequence of the implication be considered a description? as stated, only if we allow definite relations in the description language... alternatively, i think you could reify this relation using junk slots, but then you'd have to change the ontology to include "append" objects... > Of course, one can write relational constraints for 'concatenate', > if one has access to something like Prolog (or the definite clause > component of ALE). Do we want to assume that such a component is > inherently part of the underlying formalism? I think we'd better do > without it. Indeed, some pre-defined relations (say, 'append', > 'set-union' etc.) might be required, but not the full mechanism of > some logic programming language. good point... even more problematic, in my opinion, is the widespread use of relations in the *antecedents* of conditionals... that is, suppose you had a constraint like this (i'm not endorsing this constraint; it's just an example): if X is a sign, and the CONTENT value of a member Y on X's ARG-S list is npro then Y is o-free now, if you code this up using definite relations, you end up with a member relation in the antecedent of a conditional... but now in the semantics you're faced with the problem of saying what it means (denotationally) to have a *negated* member relation... and i don't think a junk slot encoding solves the problem... rather, if you go the junk slot route, i think you'd have to, for every relation, define a corresponding negated relation (not_member, in this case) and make sure that whenever the original relation appears within the scope of negation, you rewrite the formula without negations and use the corresponding negated relation... interestingly, king showed that you can eliminate negation and implication from the description language entirely if you have path inequalities and adopt the closed world assumption... this corresponds to the idea above, i think: path equalities are a kind of relation, and path inequality is the corresponding negated relation... frank richter and manfred sailer, at universit"at t"ubingen, have done some very interesting work regarding the extension of SRL with definite relations, and i should probably let them take over the discussion from here... mike calcagno
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:28 PST