Re: How do RDF and Formal Logic fit together?

> From: "Thomas B. Passin" <tpassin@home.com>
> 
> > [Seth Russell]
> >
> > > From: "Pat Hayes" <phayes@ai.uwf.edu>
> > >
> > > > I agree that would be great, but unfortunately RDF isn't quite good
> > > > enough for that. Its just *too simple* to be useable as a general
> > > > syntax model. If it had used quadruples instead of triples, or had
> > > > some kind of context or scoping mechanism, or had some way to string
> > > > together sequences without forcing the use of reification; any of
> > > > those would have worked; but plain graphs just don't cut the mustard.
> > >
> > > Well the pentuples of a mentograph would do the job, me thinks:
> > >
> > > 1) subject
> > > 2) property
> > > 3) object
> > > 4) statement ID
> > > 5) sequence
> > >
> > > ... which when you add some other needed typing info to make the data
> > > processing practical it ends up being a 7-tuple  see
> > > http://robustai.net/mentography/SemStructure3.gif   But you can still
> draw
> > > them as labeled directed pseudographs with an optional new sequence
> > > attribute labeling the arcs:
> http://robustai.net/mentography/sequence.gif
> > >
> >
> > You know, when you construct a computer model of an RDF graph, it's
> > practically impossible to do without having a triple be some kind of
> object
> > or entity.  It's a row in a database, or an edge definition showing source
> > and target, or a tuple (subject, predicate, object), or something that
> gives
> > an (local) identity to each statement.
> 
> Yes, definitely.  I think most implementations of RDF actually have this
> identifier already; all they need to do is expose it to the user.
> 
> >Surely it wouldn't be much of a step
> > to generalize that in the model and specify a way to map from the
> inevitable
> > local identifier to a globally unique URI.
> 
> Personally I don't think think it's necessary to have a globally unique ID
> for each triple and it may actually be misleading.   A triple only has
> meaning within a context.  If I assert the triple {:Goor :won :Election2000}
> it has a totally different meaning than if the US Electorical College had
> asserted that same triple.  I suppose there are context independant triples
> ... but I haven't personally run into any yet ... have you?   I think the ID
> of a triple should be stamped locally by the person reading or writing the
> triple within some context.

Well, these are two different kinds of IDs. An ID uniquely identifying
a triple <s,p,o> in the worst case is the whole triple written as a
term f(s,p,o) or as a constant s_p_o (or something like that), if you
do not want to introduce nested terms. (If you have axioms, which tell
you, that for certain kinds of triples a subset of <s,p,o> uniquely
identifies the triple, you can use that subset.) This is what we do in
O-Telos-RDF.

Once you have this ID, you can then use it to directly reference that
triple, for example <s_p_o,asserted_by,russell> and/or
<s_p_o,asserted_by,college>, without using the kind of reification
introducing additional reification triples which is included in the
current RDF version. If you want to have this author/context
information directly included in each tuple, you have to include a
fifth argument "context", which some people use, too.

> 
> Seth Russell
> 

Wolfgang

Received on Sunday, 14 October 2001 05:04:15 UTC