Relationships, context WAS: (SWAG: Re: & kindaLike that too) AND (RE:&)

[Manos Batsis <m.batsis@bsnet.gr>]

I hope you guys don't mind about joining the two threads :-)

> -----Original Message-----
> From: Sean B. Palmer [mailto:sean@mysterylights.com]

[...] At the moment, you can't introduce cyclic
> properties/classes (RDF Schema won't allow it), but once 
> those are allowed
> (it's on the RDFS issues list), it'll allow RDFS to imply 
> equivalences in a
> way that will baffle some non-logicians. Asserting that:-
> 
>    :x rdfs:subPropertyOf :y .
>    :y rdfs:subPropertyOf :z .
>    :z rdfs:subPropertyOf :x .
> 

Very glad you provided this example. At first I went like "Whoa. How am
I gonna use that?" but it may fit perfectly with a question I wanted to
ask.
I was wondering if I can use (or implement something like) what I
thought of as "Property context". For example, I would like to state
that 

:x synonumTo :y .

but using a higher resource as a context (OR, make it conditional).
Further more, I would like to extend this to have the ability to build
upon the model on document level, that is, having all my schemas etc in
a taxonomy repository or something, but also have the ability to easily
build upon the taxonomy models in document level, making them reusable
instead of static. This resembles human behavior, as all of our
thoughts, feelings, judgment etc. are always in a rather liquid state
based on time/situation etc.
Is something like this easy/currently implemented?


> Anyway, what was I talking about? Oh yes, it's quite 
> difficult to define
> what a "non-hierarchial relationship" is, because all triples 
> are miniature
> hierarchies.

I don't really see it that way. The only thing a triple gives in my head
is a dimension in a virtual space (something I don't even find relevant
but since we are talking about models...). Also, different hierarchies
can be build with the same statements, depending from the starting point
(your cyclic statements above provide an excellent example) so the
relationship as a whole cannot be resembled as one hierarchy.


> When we link them together, they form a Web, 
> which transcends
> the triples, much like the triples transcend the nodes.

Yes.

 
> -----Original Message-----
> From: Aaron Swartz [mailto:aswartz@swartzfam.com]

> 
> On Monday, August 27, 2001, at 08:48  PM, Sean B. Palmer wrote:
> 
> > I'm not too keen on "kindaLike" because all it does is state 
> > that there is
> > some level of equivalence between a subject and an object, but 
> > it doesn't
> > state the level of that equivalence. That might be useful to a 
> > human (which
> > is why I used it in the prose), but it's not going to be useful to a
> > machine, which is what the goal of the Semantic Web is. To get 
> > machines to
> > do all of the dirty work.
> 
> Actually that's not completely true. Montonic systems (or is it 
> non-monotonic... I can never remember) may use kindaLike to make 
> tentative inferences. It's rather useful for a large number of 
> applications.
> 
> "I see you're buying this book... these other books are kinda 
> like that one, and may also be interesting to you."

I think that it is rather a subject of using the ontology properly (as
html was used improperly). This "kindaLike" would be an ideal solution
for implementing dynamic navigation systems (a problem I'm facing right
now) OR, dynamic resolution of relationships by machines (if you
establish factors and make the property meaning dependent on them, such
as the "context" I am asking for above).

Kindest regards,

Manos