Re: Extra AWWSW on Tues 2 June 9:00 AM EST for IRW ontology

On Tue, Jun 2, 2009 at 8:02 AM, Harry Halpin <hhalpin@ibiblio.org> wrote:
> Looking over this diagram, I think the things missing from IRW are:
>
> -Entity: Perhaps superclass of irw:WebRepresentation
Example of an RFC 2616 entity that's not an irw:WebRepresentation, please?
(Note that Roy has asserted that all entities are representations of
some resource; just
not necessarily the resource that was requested.)

> -Generic Resource: Perhaps owl:equivalentClass irw:InformationResource
That's what Tim says, but the definitions are completely different, so a theorem
would be needed to link them.

> -Fixed Resource: rdfs:subClassOf InformationResource?
Yes, according to Tim

> - just import most of ont.rdf from TimBL, but add in our media-type
> ontology to help out with cardinality restrictions (i.e. a fixed
> resource can have exactly one media type)
> -Maybe Version: Not sure. If this a property or a class?
Sorry, doesn't genont say? I'm pretty sure it's a subclass of
generic resource (genont:InformationResource).

> - import all the time-specific/language-specific etc. directly into IRW
> - maybe awww:message rdfs:equivalentClass ire:InformationRealization
> - descriptionResource owl:equivalentClass ldow:AssociatedDescription

I don't think descriptionResource is a useful class. If you need it just
talk about the class of generic resources that are in the domain of
'describes'.

> I think we should *not* model any of the "ftrr" work, characteristic,
> or have lots of different versions of "representation", and so on. I'm
> not sure there's a difference that makes a difference with some of
> these terms, and the Boothian "ftrr" idea, while interesting, is
> incorrect and should not be modellled IMHO.

I do not think it is incorrect, as a definition cannot be incorrect;
and anyone should be free to model anything they like as far as I'm
concerned. The question is whether in any particular collaborative
ontology authoring effort, the authors *choose* to include one entity
and not another. Maybe we will choose to have "ftrr" or "trace" in the
final thing we make, or maybe not. We shall see. Your "should not" is
noted...

If you really believe in the AWWW definition of "information resource"
then I think you are obligated to model characteristics,
essentialness, conveyance and so on so that what you mean becomes
clear. However I have not heard anyone in our group stand up for that
definition. I give it air time because, and only because, it borders
on being normative. (And I give RFC 2616's definition of "resource"
prominence because it *is* normative.) I predict that in the end we
will just strategically ignore it, but this remains to be seen.

I think that if you see the diagram as a draft ontology issued by the
group you are reading way too much in it. No one but me is to blame
for it; I don't believe it myself; and I said only that we use it for
"target practice". I expect that the ultimate product of the group, if
any, will omit much of what is in the diagram; but for now the
diagram's purpose is to surface issues and attempt clarity and
comparability of warring models.

(This is a problem-solving technique I use a lot: Find a bigger
picture into which the warring factions fit, so that they can be seen
as not warring, and can be compared on their merits.)

For example, the notions of "generic resource" and "representation"
still make little sense to me ontologically and I have no confidence
I'd be able to use them or relate them to other ontologies without
violating their intent (because I still don't understand their
intent). (They still look like post hoc idealizations of the web to
me.) I am attempting to pursue them on their own terms, because I want
to hear and understand the contributions of others in the group; not
because I have any confidence in them yet.

We are not nearly as far along as you seem to think we are! We're
pretty far from consensus on anything. Please don't read too much into
our artifacts, and especially do not attribute any of them to the
group until we've reached consensus.

> It's more difficult thinking about "state". We could have a resource
> have a state, and then say the WebRepresentation represents the state,
> but then that's trying to put some temporality into an ontology (as
> the state should have some time interval), and ontologies really
> aren't that good at modelling things that change over time, and
> tracking changes is difficult if not a generally bad idea for OWL
> ontologies.

Modeling change, to the limited extent of distinguishing time-varying
from non-time-varying generic resources, is a central goal of Tim's
ontology. It bites off such a modest part of the problem that I think
it is tractable.

Quine (in "Empirical Content", if I've read it correctly) points out
that time needn't have any special status in accounting for the way we
talk about empirical theories. That is, you can move anything you want
to say that is sensitive to, or relates to, time into your observation
statements ("observation categoricals") and then just treat time in
your theory (or model), just as you would any other variable. This is
exactly what genont does.

Jonathan

> On Tue, Jun 2, 2009 at 12:22 AM, Jonathan Rees <jar@creativecommons.org> wrote:
>> Harry, for reference, the diagram we've been using for target practice at
>> our telecons is here (temporarily at least):
>>
>> http://sw.neurocommons.org/tmp/jar-diagram-5.pdf
>>
>> It was on the ESW wiki but seems to have been blown away. Will pursue fixing
>> this separately.
>>
>> The focus has been on how to write an ontology that has a chance of
>> interoperating
>> with others, so issues such as curation practice and falsifiability
>> are more important
>> than they would be in a pure linked data approach.
>>
>> Jonathan

Received on Tuesday, 2 June 2009 12:48:15 UTC