Hi Richard,
On 29/7/2014 12:06 ??, Richard Light wrote:
Martin,
Thank you for this: I now have a much better idea of what you are
trying to express. I can also now see how the AAT is relevant to the
discussion: it is precisely a set of propositions with an identity, or
(to use SKOS terminology) a Concept Scheme. (However, I can't see the
AAT being cited as justification for an assertion or set of
assertions, so maybe it's not that pertinent an example.)
However, the AAT as a concept scheme is identified by the URI:
http://vocab.getty.edu/aat/
which yields a web page when invoked normally, and redirects to:
http://vocab.getty.edu/download/rdf?uri=http://vocab.getty.edu/aat/
when RDF is requested in the HTTP Accept header. Every concept within
AAT contains an assertion that it is skos:inScheme AAT.
So, in what way would you create a Named Graph (in your sense) for the
AAT? What URI would you associate with each triple in the concept
scheme?
http://vocab.getty.edu/aat/
And what practical benefit does this give you, that simply using the
URL quoted above doesn't give you?
You can upload the AAT into a triple store and still know this is the
same AAT...
An example of a "well known" graph will always have that problem - it is
defined somewhere else.
In the CRM, we do not require examples to be well known. We can create a
small TRIG file.
The whole point of a Named Graph is distinguishing a bunch of
propositions within a database (triple store,
graph database...). If the propositions are still on the Getty server, I
cannot run a transitive closure etc. on its
content.
You can regard the whole LoD space as a big triple store, but you cannot
run a query against it up to now.
This part of LoD is still a bit in the science fiction area - at least
from the point of view of industrial application.
Distributed querying still poses theoretical questions and serious
performance issues. I'd guess some
20 years to become reality, if ever. The "if ever" comes from my
concerns how you can manage such a space
without controlled provenance of knowledge with all its ramifications.
Best,
Martin
Richard
On 28/07/2014 11:34, martin wrote:
Dear Richard,
On 28/7/2014 11:41 ??, Richard Light wrote:
Martin,
I thought that a major merit of the CRM was that it was an abstract
model, which could be instantiated using whatever technology was
felt to be appropriate. That being the case, I would be concerned
if RDF-specific techniques were presented to the world as the only
way in which a particular challenge ("implementing argumentation
systems ...") could be tackled using the CRM. Or are you talking
specifically about RDF implementations of the CRM?
I share your concerns :-) !
Why can't "premises and conclusions" be modelled using reification,
so they can then be given a unique URI? This is the sort of approach
which the BM has successfully deployed, as I understand it. I would
be grateful if someone could provide a really simple concrete
example which shows the need for the named graph approach.
Your are right!
Actually I see the "Named Graph" not as a particular RDF feature, but
at the level of abstraction that Simon pointed
out: A set of propositions with a "historical" identity which is not
reduced to the identity of the set itself.
The CRM uses an abstract data model of classes, superclasses,
properties, superproperties etc., which is more or
less the stable core of all data structures and KR models used so far
in industrial systems. We have however adopted
the term "property" from RDF, just to reduce the semantic gap for
people now. Originally, we used TELOS terms, but KIF, OIL was equally
compatible.
The requirement to introduce argumentation structures into consistent
graphs of propositions is relatively new.
Reification is an atomic mechanism, which does not allow for
describing that a set of propositions is believed
together. Therefore it looses an important part of the semantics of
argumentation. A Named Graph is in my mind
an abstarction which subsumes reification. Reification is a
workaround using a syntax which has not foreseen the problem before.
Named Graph is a NEW logical construct not found in any other
industrial KR model, and born out of a necessity that first showed up
when integrating different sources. (Before, one could say AI just
slept in a one-truth cyberworld with a god-like user or math on top
of reality).
I believe we need the Named Graph construct as a logical form, not as
an RDF syntax, if we want to integrate
provenance of knowledge with the CRM. So far, we have evidence of two
real-life data structures, one is
archaeological excavation records, and another description of
medieval book-bindings, which systematically
register source of evidence and concluded facts. E.g., geometric
topology of stratigarphic units and microsopic
stratigraphic interface properties are used to justify chronological
sequence. In a simple model, this is atomic,
in a more general, it is probabilistic Bayesian. So, we would need a
"Typed Named Graph", which restricts the
propositions in the Graph to a certain schema (topology, chronology),
and then a relationship "is evidence for"
between the typed named graphs. The assertion itself forms part of
the belief implicit in the archaeological
record.
If there is any logician on this mailing list, a proper formulation
of such a construct and an abstract syntax for the CRM would be great
to have!!!
We will try to suggest a graphic primitive, which is a bubble around
the propositions with a "hot spot" on the
perimeter.
Suggestions most welcome!
To pick up on the suggestion of using the AAT as an example: in what
way is the AAT a named graph? Surely it's a SKOS Concept Scheme
(plus)? I think it would be impossible to give an example of a
"well-known" named graph, for the reasons Simon has been explaining.
Named Graphs are new, so none is really "well known", but I would
regard a skosified AAT as a Named Graph,
as well as all the RDF junks for LoD, once RDF regards any RDF file
as a Named Graph. The only condition is, that
two RDF Files with the same content and different URI are not
regarded as being identical (owl:same_as).
Best,
Martin
--
*Richard Light*
_______________________________________________
Crm-sig mailing list
Crm-sig@ics.forth.gr
http://lists.ics.forth.gr/mailman/listinfo/crm-sig
--
--------------------------------------------------------------
Dr. Martin Doerr | Vox:+30(2810)391625 |
Research Director | Fax:+30(2810)391638 |
| Email: mar...@ics.forth.gr |
|
Center for Cultural Informatics |
Information Systems Laboratory |
Institute of Computer Science |
Foundation for Research and Technology - Hellas (FORTH) |
|
N.Plastira 100, Vassilika Vouton, |
GR70013 Heraklion,Crete,Greece |
|
Web-site: http://www.ics.forth.gr/isl |
--------------------------------------------------------------