Re: Entailment versus implication

(Sorry about late reply, getting behind with email again.)

>The recent discussions regarding reification and entailment
>have made me suspect that I am not understanding (again) the
>MT vocabulary.
>
>My understanding of entailment (coming from a computational
>linguistics background) is that a statement is entailed by a
>set of statements if it is logically implied by the set
>of statements (i.e. a triple is entailed by an RDF graph if
>it is logically implied by that RDF graph).
>
>To me, "entailment" means that adding to the graph does not
>increase the information already there, even if it makes some
>information explicit that was otherwise implicit. Everything is
>already there. E.g.

Right, exactly.

>   Club members get a discount.
>   Jill is a club member.
>
>entails
>
>   Jill gets a discount.
>
>
>So, when folks say that
>
>_:B ex:father #Bob .
>_:B ex:gender ex:Male .
>_:G ex:father #Bob .
>_:G ex:gender ex:Female .
>
>entails
>
>_:B ex:gender ex:Female .
>_:G ex:gender ex:Male .
>
>I start to wonder if we are all talking about the same thing.
>
>Certainly the first set of triples do *not* imply the latter
>pair of triples. How could they?

Well, it goes outside RDF, but one could argue that if it is known 
that ex:father is functional, ie people only have one father, then

_:B ex:father #Bob .
_:G ex:father #Bob .

together entail _:G = :_B, and then you can get the conclusion by 
substitution. THis does require that extra assumption, though, to be 
a valid entailment, and neither the assumption nor this kind of 
equality reasoning is supported by RDF. (You could do this kind of 
reasoning, expressed in terms of class cardinalities, in DAML.)

>Where is the basis for such
>an implication? The existence of some bNode?
>
>Where are we having the breakdown in communication? (or
>are we ;-)

I think that some of us are treating RDF entailment strictly in RDF, 
while others see it as embedded in a larger 'web' of entailment that 
might use hypotheses that come from outside RDF. The problem (to me) 
of the latter view is that there is no way to say what might or might 
not be entailed by anything in this case.

Pat
-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Monday, 11 February 2002 10:34:53 UTC