On Jul 26, 2010, at 11:40 AM, John Bottoms wrote: (01)
> Pat,
>
> Hmmm, if a quad store simply establishes an additional link for
> a context (or some other entity), then why could we not use the
> predicate mechanism of a triple store to access the 4th link?
> Must we also standardize on predicates?
>
> Since RDF handles n-ary graphs using blank nodes, why couldn't
> quad-graphs also be handled with blank nodes (with an appropriate
> predicate)?
>
> -John Bottoms
> Firststar Systems
> Concord, MA
>
> (misquoting...)
> “When I use a graph,” Humpty Dumpty said in rather a scornful
> tone, “it means what I choose it to mean, neither more nor
> less.” (02)
Well, that is the problem, right there. RDF graphs can't be used that
freely, since RDF also comes with a normative semantics, which, while
minimal, does have to be respected. Humpty can insist that it means
what he wants it to mean, as long as that meaning can be shoehorned
into the RDF semantics. So if we have something like CL (or indeed OWL
for that matter) which has a semantics of its own, which is itself an
extension of the RDF semantics (which indeed it is: RDF is in fact the
fragment of CL which allows conjunctions of atomic sentences of the
form (P A B), and existential quantification on the A,B names but not
on the P name - plus some way to encode literals as terms, which is
easy but which I will ignore for now) then what we want is some way to
embed CL into RDF in a way that preserves as much as possible of the
semantics, ideally which keeps the RDF being the RDF fragment of CL.
That way, simple content expressed in RDF still means, in CL/RDF, what
it would mean in plain RDF. In other words, the CL/RDF is a "semantic
extension" of RDF. (The OWL embedding into RDF has this desirable
property, for example.) (03)
To do this for CL requires some way to represent negation scopes, to
encode the difference between for example
(not (and (P A B)(Q B A)))
and
(and (not (P A B)) (Q B A))
Now, one attitude to take to this is to observe that these are all S-
expressions, which can be implemented LISP-style using paired links
which in turn can be encoded in RDF triples using, for example, the
RDF collections vocabulary (itself modeled on LISP). So of course this
can be put into a triple store. But this simply encodes the logical
*syntax*, not its semantics. If we want the CL/RDF to be an RDF
semantic extension, then the RDF that encodes the CL (04)
(and (P A B)(Q B A)) (05)
has to be the RDF graph (06)
A P B .
B Q A . (07)
and then the 'not' applied to this has to be a graph that somehow
includes this one but in such a way that it becomes negated. And this
is *impossible* in current RDF, because it would be nonmonotonic,
which is explicitly disallowed by the RDF semantics. You can't change
the RDF meaning of some RDF by adding something else to it: in
particular, you can't change it form truth to falsity, ie logically
negate it. So you have to have a syntax that goes beyond the RDF
triples model, and the simplest one is to move to is quads, which
turns out to be just enough (if you extend the semantics properly).
For details, see my invited talk on 'blogic' at the last ISWC. (08)
Pat (09)
>
> Pat Hayes wrote:
>
>>
>> On Jul 24, 2010, at 11:34 AM, Cameron Ross wrote:
>>
>>> I've been reviewing the BioPortal documentation in preparation for
>>> the
>>> CL/OOR discussion. Page 4 of the BioPortal Architecture Roadmap
>>> states that "In Year Five, we will complete the transition to a
>>> unified RDF Triple Store for all the artifacts stored in BioPortal".
>>> How do we store CL content within a triple store?
>>
>>
>> You can't, in general. You could, in principle, in a quad store, but
>> only very awkwardly and by using a bunch of conventions that havn't
>> yet
>> been standardized.
>>
>> Pat
>>
>>>
>>> see
>https://bmir-gforge.stanford.edu/gf/download/docmanfileversion/272/391/NCBOArchitectureRoadmapReport20080424_FINAL.pdf
>
>>> .
>>>
>>> Cameron.
>>> --
>>> Kojeware Corporation
>>>
>>> _________________________________________________________________
>>> Message Archives: http://ontolog.cim3.net/forum/oor-forum/
>>> Subscribe: mailto:oor-forum-join@xxxxxxxxxxxxxxxx
>>> Config/Unsubscribe: http://ontolog.cim3.net/mailman/listinfo/oor-forum/
>>> Shared Files: http://ontolog.cim3.net/file/work/OOR/
>>> Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository
>>
>>
>> ------------------------------------------------------------
>> IHMC (850)434 8903 or (650)494
>> 3973
>> 40 South Alcaniz St. (850)202 4416 office
>> Pensacola (850)202 4440 fax
>> FL 32502 (850)291 0667 mobile
>> phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
>>
>>
>>
>>
>>
>> ------------------------------------------------------------------------
>>
>>
>> _________________________________________________________________
>> Message Archives: http://ontolog.cim3.net/forum/oor-forum/
>> Subscribe: mailto:oor-forum-join@xxxxxxxxxxxxxxxx
>> Config/Unsubscribe: http://ontolog.cim3.net/mailman/listinfo/oor-forum/
>> Shared Files: http://ontolog.cim3.net/file/work/OOR/
>> Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository
>
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/oor-forum/
> Subscribe: mailto:oor-forum-join@xxxxxxxxxxxxxxxx
> Config/Unsubscribe: http://ontolog.cim3.net/mailman/listinfo/oor-
> forum/
> Shared Files: http://ontolog.cim3.net/file/work/OOR/
> Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository
> (010)
------------------------------------------------------------
IHMC (850)434 8903 or (650)494 3973
40 South Alcaniz St. (850)202 4416 office
Pensacola (850)202 4440 fax
FL 32502 (850)291 0667 mobile
phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes (011)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/oor-forum/
Subscribe: mailto:oor-forum-join@xxxxxxxxxxxxxxxx
Config/Unsubscribe: http://ontolog.cim3.net/mailman/listinfo/oor-forum/
Shared Files: http://ontolog.cim3.net/file/work/OOR/
Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository (012)
|