skos:[exactMatch|closeMatch|broader|narrower] vs owl:sameAs

18 views
Skip to first unread message

Evan Sandhaus <sandhes@nytimes.com>

unread,
Nov 3, 2009, 3:14:56 PM11/3/09
to Pedantic Web Group
Last week at ISWC, The New York Times launched data.nytimes.com.
Since then I've heard a number of concerns over our use of owl:sameAs
and suggestions that we replace our owl:sameAs predicates the
appropriate skos predicates.

In researching this issue, I've come up with a question that I'd like
to put to this group.

The domain and range of skos:[exactMatch|closeMatch|broader|narrower]
and other skos:semanticRelation(s) is skos:concept and the resources
I'm currently mapping to ( at dbpedia and freebase) are not of
rdf:type skos:concept. Thus in these instances, it seems incorrect
for me link to resources that are not of type skos:concept using
skos:semanticRelation(s) predicates.

Do I have this right? And, if so, is this a big enough concern that I
should not use the skos:semanticRelation(s) in this context?

Thanks,

Evan Sandhaus
Semantic Technologist
New York Times Research + Development

Kingsley Idehen

unread,
Nov 3, 2009, 4:18:04 PM11/3/09
to pedant...@googlegroups.com
Evan,

Take 2: you are full entitled to make co-reference claims in your data
space. If you deem identifiers X & Y as referring to the same thing (in
your world view) then by all means say so. Of course, its better if you
ensure that you co-referenced entities are of the same Type, so a
Document != Person etc..

The semantics of owl:sameAs are quite unambiguous, no where is the
definition is Transitivity or Symmetry inferred or implied as part of
the definition of this property. It is exactly as defined in the spec, a
mechanism for co-reference assertions by the owner of a Data Space.

Reasoning over you assertions, like all things in life, is a purely
subjective act on the part of the entity (human or agent) that seeks to
reason based on your assertions. Of course, clarity helps.

The label "owl:sameAs" cannot override the actual description of the
property. Strangely, people remain fixated on the "label" rather than
the formally defined semantics [1][2] :-(

To conclude, reasoning is an optional and subjective act, co-references
between entities of the same type (is desired and very nice to have),
you are totally entitled to your "world view" on the Web just as you are
in real life, of course quality matters, but being "nearly pregnant" is
the kind of quality Linked Dta needs :-)

Links:

1.
http://linkeddata.uriburner.com/about/html/http/www.w3.org/2002/07/owl%01sameAs
2. http://www.w3.org/TR/owl-ref/#sameAs-def


--


Regards,

Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO
OpenLink Software Web: http://www.openlinksw.com


Richard Cyganiak

unread,
Nov 3, 2009, 6:56:55 PM11/3/09
to pedant...@googlegroups.com
Hi Kingsley,

Is it just me or are you selling something?

This is a list for spec lawyering, not for evangelism. Please tread
lightly and consider well when posting here. Most of all, read and
understand the question before answering. Finally, make sure you have
your facts straight. owl:sameAs is symmetric and transitive. This
follows from the Direct Model-Theoretic Semantics for OWL [1].
However, I don't know what that has to do with Evan's question, which
was about applying SKOS properties to things other than skos:Concepts.

That being said, welcome to the list. Again, please tread lightly.
This is not the right forum to advance particular worldviews on how to
get the most value out of SemWeb technologies, no matter how right you
might be.

All the best,
Richard

[1] http://www.w3.org/TR/owl-semantics/direct.html#3.3

Kingsley Idehen

unread,
Nov 3, 2009, 9:20:16 PM11/3/09
to pedant...@googlegroups.com
Richard Cyganiak wrote:
>
> Hi Kingsley,
>
> Is it just me or are you selling something?
I think its just you re. this matter :-)

>
> This is a list for spec lawyering, not for evangelism. Please tread
> lightly and consider well when posting here. Most of all, read and
> understand the question before answering. Finally, make sure you have
> your facts straight. owl:sameAs is symmetric and transitive.

> This follows from the Direct Model-Theoretic Semantics for OWL [1].

Where does it state that its Symmetric and Transitive? The property has
a URI.

My fundamental point: should we be arguing about the meaning of
something in prose when it has a URI that should convey its exact semantics?

Example:
http://linkeddata.uriburner.com/about/html/http/www.w3.org/2004/02/skos/core%01narrowerTransitive
. Compare that to:
http://linkeddata.uriburner.com/about/html/http/www.w3.org/2002/07/owl%01sameAs


As you can see here is the basic problem: one thing preached (stuff all
that matters in structured form behind a URI) but a totally different
thing in reality. The debate about "owl:sameAs" would be best served by
its URI conveying its semantics unambiguously.


> However, I don't know what that has to do with Evan's question, which
> was about applying SKOS properties to things other than skos:Concepts.

Well, if you look at what Evan is seeking, its about alternatives to his
use of owl:sameAs, at least that's how I read his comments.


Kingsley

>> co-references between entities of the same type (is desired and very

Evan Sandhaus <sandhes@nytimes.com>

unread,
Nov 3, 2009, 10:44:01 PM11/3/09
to Pedantic Web Group
Forgive me if I sound ... um ... pedantic, but I'm still curious
about the question I raised previously.

Namely:

The domain and range of skos:[exactMatch|closeMatch|broader|narrower]
and other skos:semanticRelation(s) is skos:concept and the resources
I'm currently mapping to ( at dbpedia and freebase) are not of
rdf:type skos:concept. Thus in these instances, it seems incorrect
for me link to resources that are not of type skos:concept using
skos:semanticRelation(s) predicates.

The question is, am I just trading a big infrencing problem for a
smaller one?

Thanks,

Evan

On Nov 3, 9:20 pm, Kingsley Idehen <kide...@openlinksw.com> wrote:
> Richard Cyganiak wrote:
>
> > Hi Kingsley,
>
> > Is it just me or are you selling something?
>
> I think its just you re. this matter :-)
>
>
>
> > This is a list for spec lawyering, not for evangelism. Please tread
> > lightly and consider well when posting here. Most of all, read and
> > understand the question before answering. Finally, make sure you have
> > your facts straight. owl:sameAs is symmetric and transitive.
> > This follows from the Direct Model-Theoretic Semantics for OWL [1].
>
> Where does it state that its Symmetric and Transitive? The property has
> a URI.
>
> My fundamental point: should we be arguing about the meaning of
> something in prose when it has a URI that should convey its exact semantics?
>
> Example:http://linkeddata.uriburner.com/about/html/http/www.w3.org/2004/02/sk...
> . Compare that to:http://linkeddata.uriburner.com/about/html/http/www.w3.org/2002/07/ow...
> >>http://linkeddata.uriburner.com/about/html/http/www.w3.org/2002/07/ow...
>
> >> 2.http://www.w3.org/TR/owl-ref/#sameAs-def

Rinke Hoekstra

unread,
Nov 4, 2009, 2:58:33 AM11/4/09
to pedant...@googlegroups.com
Hi,

On 4 nov 2009, at 03:20, Kingsley Idehen wrote:
>> This follows from the Direct Model-Theoretic Semantics for OWL [1].
>
> Where does it state that its Symmetric and Transitive? The property
> has a URI.

As Richard says, it is in the spec. The URI of the property has
nothing to do with it, other than that it indicates that the sameAs
property is defined within the OWL namespace...

> My fundamental point: should we be arguing about the meaning of
> something in prose when it has a URI that should convey its exact
> semantics?

URI's don't convey semantics, formal specifications do.

The URL above does not specify the semantics of owl:sameAs, it is
'just' a non-normative RDF schema description of the primitives of
OWL. It was never meant as a normative reference, and IMHO it is
wildly confusing.

Nonetheless, the URI should redirect to the new OWL 2 compliant
owl.owl file as specified by the OWL WG. I will take this up with the
responsible persons.

> As you can see here is the basic problem: one thing preached (stuff
> all that matters in structured form behind a URI) but a totally
> different thing in reality. The debate about "owl:sameAs" would be
> best served by its URI conveying its semantics unambiguously.

The new owl.owl file does specify owl:sameAs as transitive and
symmetric. Even so, one should refer to the normative documents when
digging for the semantics of OWL 2 primitives.

Cheers,

Rinke


---
Dr Rinke Hoekstra

AI Department | Leibniz Center for Law
Faculty of Sciences | Faculty of Law
Vrije Universiteit | Universiteit van Amsterdam
De Boelelaan 1081a | Kloveniersburgwal 48
1081 HV Amsterdam | 1012 CX Amsterdam
+31-(0)20-5987752 | +31-(0)20-5253499
hoek...@few.vu.nl | hoek...@uva.nl

Homepage: http://www.few.vu.nl/~hoekstra


Rinke Hoekstra

unread,
Nov 4, 2009, 3:00:46 AM11/4/09
to OWL 2, pedant...@googlegroups.com
Hi all,

An issue that came up on the Pedantic Web mailinglist: the OWL
namespace URI does not (yet) redirect to the new owl.owl file as
specified by Michael.

Best,

Rinke

Begin forwarded message:

Rinke Hoekstra

unread,
Nov 4, 2009, 3:07:06 AM11/4/09
to pedant...@googlegroups.com, Antoine Isaac
Hi Evan,

Perhaps Antoine is a better person to explain this, but here is my try.

With respect to your skos:semanticRelation question: it is perfectly
ok to use SKOS relations between your resources, as long as you don't
mind reasoners inferring the type of these resources to be
skos:Concept. IMO this propagation of types to external repositories
(dbpedia, freebase) is far less harmful than the implications of using
owl:sameAs relations. For all practical purposes, dbpedia resources
*are* concepts.

The domain and range restrictions on SKOS relations do not function as
integrity constraints, there is no *requirement* for the relata to be
of type skos:Concept. For a short explanation, see [1].

Best,

Rinke

[1] http://www.w3.org/TR/2009/REC-skos-reference-20090818/#L881

Alexandre Passant

unread,
Nov 4, 2009, 3:08:05 AM11/4/09
to pedant...@googlegroups.com
Hi Evan,

On 4 Nov 2009, at 03:44, Evan Sandhaus <san...@nytimes.com> wrote:

>
> Forgive me if I sound ... um ... pedantic, but I'm still curious
> about the question I raised previously.
>
> Namely:
>
> The domain and range of skos:[exactMatch|closeMatch|broader|narrower]
> and other skos:semanticRelation(s) is skos:concept and the resources
> I'm currently mapping to ( at dbpedia and freebase) are not of
> rdf:type skos:concept. Thus in these instances, it seems incorrect
> for me link to resources that are not of type skos:concept using
> skos:semanticRelation(s) predicates.

When you will use skos:xxx (with domain skos:Concept) to your resource
X, you don't need to have previously defined X as a skos:Concept.
However, as soon as you use that property, X will become a skos:Concept.
This is based on the RDFS semantics [1] and its entailment rules [2],
namely rule rdfs2

So the question is then, is that a problem if X is both a skos:Concept
and, e.g., a foaf:Person ?

AFAIK, there is currently no formal disjunction between both in the
respective ontologies, so that will not lead to consistency issues.
However, from my point of view, skos:Concept should be used when
dealing with abstract notions rather than physical objects.
But the SKOS spec simply says "A SKOS concept can be viewed as an idea
or notion; a unit of thought. However, what constitutes a unit of
thought is subjective, and this definition is meant to be suggestive,
rather than restrictive."

So, that's a matter of subjectivity, but will no harm theoretically
since there is no disjunction between both.
BTW, Richard launched a thread on the topic on the dbpedia-discussion
and the SKOS list, so we may have clarification from the SKOS guys.

Hope that helps,

Alex.

[1] http://www.w3.org/TR/rdf-mt/
[2] http://www.w3.org/TR/rdf-mt/#RDFSRules

--
Dr. Alexandre Passant
Digital Enterprise Research Institute
National University of Ireland, Galway
:me owl:sameAs <http://apassant.net/alex> .


Rinke Hoekstra

unread,
Nov 4, 2009, 3:10:46 AM11/4/09
to pedant...@googlegroups.com
Thanks Alexandre, for explaining this far more eloquently than I did.

-Rinke

Rinke Hoekstra

unread,
Nov 4, 2009, 5:41:28 AM11/4/09
to pedant...@googlegroups.com
Hi Pedants,

A slight correction:

On 4 nov 2009, at 09:00, Rinke Hoekstra wrote:
>> The new owl.owl file does specify owl:sameAs as transitive and
>> symmetric.

Actually it doesn't.

>> Even so, one should refer to the normative documents when digging
>> for the semantics of OWL 2 primitives.

See the comment field on the Ontology element (note the "SHOULD NOT"):

This ontology partially describes the built-in classes and
properties that together form the basis of the RDF/XML syntax of
OWL 2.
The content of this ontology is based on Tables 6.1 and 6.2
in Section 6.4 of the OWL 2 RDF-Based Semantics specification,
available at http://www.w3.org/TR/owl2-rdf-based-semantics/.
Please note that those tables do not include the different
annotations
(labels, comments and rdfs:isDefinedBy links) used in this file.
Also note that the descriptions provided in this ontology do not
provide a complete and correct formal description of either the
syntax
or the semantics of the introduced terms (please see the OWL 2
recommendations for the complete and normative specifications).
Furthermore, the information provided by this ontology may be
misleading if not used with care. This ontology SHOULD NOT be
imported
into OWL ontologies. Importing this file into an OWL 2 DL ontology
will cause it to become an OWL 2 Full ontology and may have other,
unexpected, consequences.

Best,

Rinke

Kingsley Idehen

unread,
Nov 4, 2009, 6:40:48 AM11/4/09
to pedant...@googlegroups.com
Rinke Hoekstra wrote:
>
> Hi,
>
> On 4 nov 2009, at 03:20, Kingsley Idehen wrote:
>>> This follows from the Direct Model-Theoretic Semantics for OWL [1].
>>
>> Where does it state that its Symmetric and Transitive? The property
>> has a URI.
>
> As Richard says, it is in the spec. The URI of the property has
> nothing to do with it, other than that it indicates that the sameAs
> property is defined within the OWL namespace...
>
>> My fundamental point: should we be arguing about the meaning of
>> something in prose when it has a URI that should convey its exact
>> semantics?
>
> URI's don't convey semantics, formal specifications do.
Rinke,

>
>> Example:
>> http://linkeddata.uriburner.com/about/html/http/www.w3.org/2004/02/skos/core%01narrowerTransitive .
>> Compare that to:
>> http://linkeddata.uriburner.com/about/html/http/www.w3.org/2002/07/owl%01sameAs
>>
>
> The URL above does not specify the semantics of owl:sameAs, it is
> 'just' a non-normative RDF schema description of the primitives of
> OWL. It was never meant as a normative reference, and IMHO it is
> wildly confusing.
Basically the essence of the matter!

>
> Nonetheless, the URI should redirect to the new OWL 2 compliant
> owl.owl file as specified by the OWL WG. I will take this up with the
> responsible persons.

Great!


>
>> As you can see here is the basic problem: one thing preached (stuff
>> all that matters in structured form behind a URI) but a totally
>> different thing in reality. The debate about "owl:sameAs" would be
>> best served by its URI conveying its semantics unambiguously.
>
> The new owl.owl file does specify owl:sameAs as transitive and
> symmetric. Even so, one should refer to the normative documents when
> digging for the semantics of OWL 2 primitives.

An HTTP URI of a property should convey its semantics (if such are
fully expressible and if not, there are annotation slots for provide
further guidance in prose). FWIW - I am being deliberately pedantic.

I my world view HTTP URIs are supposed enable us discern descriptions of
their referents, in a structured and unambiguous manner, when such is
possible. In this case fixing the description of the "owl:sameAs"
property, in a manner that's discernible via its HTTP URI is all that
was required to close the matter, which is the case right now :-)


Kingsley

>>>> co-references between entities of the same type (is desired and

Kingsley Idehen

unread,
Nov 4, 2009, 6:49:34 AM11/4/09
to pedant...@googlegroups.com
Rinke,

Contrary to some very strange misconceptions. I spend a hell of a lot of
time on these matters before I deem it worthy of my time to make a post.

As you can see, it's a mess.

The whole OWL gig was supposed to be about: "Say What You Mean" so that
its much easier to: "Mean What You Say", with machines processors in
mind. It was never supposed to be about prose for humans to debate
about, bar those matters that aren't expressible in OWL.

I use owl:sameAs subjectively when reasoning on it. And that's all that
people can do right now. My rule of thumb is not to assert owl:sameAs
links between resources of different types, or types that are explicitly
asserted to be disjoint.

Kingsley

>>>>>> co-references between entities of the same type (is desired and

>>>>>> very nice to have), you are totally entitled to your "world view"
>>>>>> on the Web just as you are in real life, of course quality
>>>>>> matters, but being "nearly pregnant" is the kind of quality
>>>>>> Linked Dta needs :-)
>>>>>>
>>>>>> Links:
>>>>>>
>>>>>> 1.
>>>>>> http://linkeddata.uriburner.com/about/html/http/www.w3.org/2002/07/owl%01sameAs
>>>>>>
>>>>>> 2. http://www.w3.org/TR/owl-ref/#sameAs-def
>>>>>>
>>>>>>
>>>>>> --
>>>>>>
>>>>>>
>>>>>> Regards,
>>>>>>
>>>>>> Kingsley Idehen Weblog:
>>>>>> http://www.openlinksw.com/blog/~kidehen
>>>>>> President & CEO OpenLink Software Web: http://www.openlinksw.com
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>>
>>>>
>>>> Regards,
>>>>
>>>> Kingsley Idehen Weblog:
>>>> http://www.openlinksw.com/blog/~kidehen
>>>> President & CEO OpenLink Software Web: http://www.openlinksw.com
>>>>
>>>>
>>>>
>>>>
>>>
>>>

Hogan, Aidan

unread,
Nov 4, 2009, 10:37:59 AM11/4/09
to pedant...@googlegroups.com
Hi Kingsley,

> The whole OWL gig was supposed to be about: "Say What You Mean" so
that
> its much easier to: "Mean What You Say", with machines processors in
> mind. It was never supposed to be about prose for humans to debate
> about, bar those matters that aren't expressible in OWL.
>
> I use owl:sameAs subjectively when reasoning on it. And that's all
that
> people can do right now. My rule of thumb is not to assert owl:sameAs
> links between resources of different types, or types that are
explicitly
> asserted to be disjoint.

As already mentioned, the semantics of owl:sameAs are already
normatively defined. The term is a built-in (much like rdf:type)
intended to represent equality, and, as such, has special treatment. The
URI need not be rigorously defined since any system supporting OWL will
already have the semantics of the term built-in (just like rdf:type):
whether or not owl:sameAs is defined as an owl:TransitiveProperty or
owl:SymmetricProperty does not matter.

The complete semantics of the OWL built-ins cannot be expressed in OWL
itself: you should refer to the normative semantics and not the owl.owl
doc. You understand the semantics of rdf:type as indicating class
membership even though the semantics attached to the URI do not indicate
this. Why don't you understand the semantics of owl:sameAs in it's
status as a built-in representing an equality relation? Would you use
rdf:type to relate any resource to any class happily knowing that you're
sticking to the domain (rdfs:Resource) and range (rdfs:Class) of
rdf:type as defined in the rdfs.rdfs doc?

Hence, you should only use owl:sameAs as if it is transitive or
symmetric, and not with the level of "subjectivity" I understand you to
mean. Reasoners handle owl:sameAs relations, and most reasoners (be they
off-the-shelf DL reasoners, rule based reasoners, or home-made custom
systems) will interpret owl:sameAs as transitive and symmetric.

For example, look at how the OWL 2 RL ruleset supports equality through
owl:sameAs [1, Table 4]. Reasoners implementing such a ruleset do not
care in the least whether the built-in owl:sameAs property is defined as
owl:TransitiveProperty or owl:SymmetricProperty. You should use
owl:sameAs in such a way that rules eq-sym, eq-trans, eq-rep-s, eq-rep-p
and eq-rep-o will, in your view, produce useful and correct inferences.
Other reasoning systems will produce similar results. This is the reason
why you assert owl:sameAs relations.

Again, the semantics implied by rules "eq-rep-s", "eq-rep-p" and
"eq-rep-o" cannot be defined in terms of OWL for owl:sameAs in the
owl.owl doc. Does this mean that such semantics should be just ignored?

Hope this clarifies,
Aidan

[1] http://www.w3.org/TR/owl2-profiles/#OWL_2_RL

Evan Sandhaus <sandhes@nytimes.com>

unread,
Nov 4, 2009, 10:42:39 AM11/4/09
to Pedantic Web Group
Alex,

Thanks so much for the clarification. So - if I understand then -
RDFS semantics are akin to what a coder like me would call "duck"
typing. That is to say, if an resource looks and acts like a type an
inferencer will deduce it to be of that type. Based on your feedback
and the SKOS definition, I think it wise to avoid using skos
predicates when mapping from NYT (people, place, location or title)
resources to DBPedia and Freebase as their resources might not
strictly be "idea[s] or notion[s]; [or] a unit of thought." That
being said, if I declare my skos:concept to be owl:sameAs a these
resources, I will essentially be throwing an inferencer down the same
rabbit hole - but I suppose that's inevitable.

I will however, use skos predicates when mapping our "descriptor"
resources as these seem to fit more comfortably into the skos
paradigm. I will also use skos predicates should we map to Library of
Congress subject headings as they are also published as
skos:resources.

Thanks again everybody for your help on resolving this issue,

Evan

Rinke Hoekstra

unread,
Nov 4, 2009, 10:42:55 AM11/4/09
to pedant...@googlegroups.com
Hi Aidan,

Thanks for that!

Rinke

Kingsley Idehen

unread,
Nov 4, 2009, 11:32:11 AM11/4/09
to pedant...@googlegroups.com
Are you saying that the definition of owl:sameAs cannot simply expose
the annotations above?

For what its worth (dare I be accused of selling something). My product
handles owl:sameAs transitivity fine, but as a matter of choice for the
Virtuoso user. I really don't need any kind of lesson on Transitivity or
how to read specs.

All I really want is for us to dog-food a little better. Pack what you
can into a URI. Annotate what isn't in there. If we adhere to the very
practices we preach to others, there will be far less confusion.

btw - How come skos:narrowerTransitive has a fuller description? I
deliberately juxtaposed skos:narrowerTransitive and owl:sameAs in my
response. Is it really problematic for the owl:sameAs URI to match the
descriptive fidelity exposed by the skos:narrowerTransitive property
URI? Wouldn't that save everyone a lot of time?


> For example, look at how the OWL 2 RL ruleset supports equality through
> owl:sameAs [1, Table 4]. Reasoners implementing such a ruleset do not
> care in the least whether the built-in owl:sameAs property is defined as
> owl:TransitiveProperty or owl:SymmetricProperty.

Hmm. really. And you really think that's how it should be done i.e,
don't care about the definition of the property, just crack along etc..
So what purpose do "owl:TransitiveProperty" and
"owl:SymmetricalProperty" serve? How about: I make those assertions in
my own named graph and then use that as my rules context basis for a
specific interaction with the data in my own data space, one which may
or may not be published to the public Web.


> You should use
> owl:sameAs in such a way that rules eq-sym, eq-trans, eq-rep-s, eq-rep-p
> and eq-rep-o will, in your view, produce useful and correct inferences.
> Other reasoning systems will produce similar results. This is the reason
> why you assert owl:sameAs relations.
>

Again, I simply don't need an "owl:sameAs" lecture. I have lots of live
demonstrations about the virtues of owl:sameAs that leverage its
transitivity (scoped to the context within which my demos operate, as
per comments above). I can share the links with you should not be able
to find them etc..


> Again, the semantics implied by rules "eq-rep-s", "eq-rep-p" and
> "eq-rep-o" cannot be defined in terms of OWL for owl:sameAs in the
> owl.owl doc. Does this mean that such semantics should be just ignored?
>

BTW - I never once even made reference to OWL2 in this entire
conversation, I don't recall any link that resolved to an OWL doc etc..

> Hope this clarifies,
>
Sorry, but I didn't need a lecture. My issue is more to do with pedantic
dog-fooding.

> Aidan
>
> [1] http://www.w3.org/TR/owl2-profiles/#OWL_2_RL

Alex Tucker

unread,
Nov 4, 2009, 11:51:26 AM11/4/09
to pedant...@googlegroups.com
Kingsley, Aiden, Rinke, etc.

Isn't the issue here that we want our ontologies and RDF to be as self
describing as possible, but when we lift up into the language we use to
make those descriptions in the first place, OWL, we can't easily make
the language constructs themselves self describing.

Ok, so OWL can't bootstrap itself, but this isn't really the point, is
it? How could we simply describe for example owl:TransitiveProperty to
a machine without a lot more machinery than we need?

Isn't this just a corner case that isn't worth the expense of worrying
about, in the same way that no-one would want to write a Javascript
interpreter in Javascript (maybe a bad example, but you get the drift)?

Alex.

Kingsley Idehen

unread,
Nov 4, 2009, 12:27:28 PM11/4/09
to pedant...@googlegroups.com
Alex Tucker wrote:
>
> Kingsley, Aiden, Rinke, etc.
>
> Isn't the issue here that we want our ontologies and RDF to be as self
> describing as possible, but when we lift up into the language we use
> to make those descriptions in the first place, OWL, we can't easily
> make the language constructs themselves self describing.
Yes, and when we hit this snag we can put the prose in an annotation.

>
> Ok, so OWL can't bootstrap itself, but this isn't really the point, is
> it? How could we simply describe for example owl:TransitiveProperty
> to a machine without a lot more machinery than we need?
But simply have the triple in the description of the property doesn't
break anything, or does it? Adding prose where semantics can't be
expressed etc.. via annotations also works. End product: a URI that
works for Humans or Machines depending on interaction context.

>
> Isn't this just a corner case that isn't worth the expense of worrying
> about, in the same way that no-one would want to write a Javascript
> interpreter in Javascript (maybe a bad example, but you get the drift)?
I am simply seeking triples for Symmetry and Transitivity in the
definition of the owl:sameAs property + annotations for those things
that can't be expressed in OWL. Then I have one URI that tells the story
in a manner that works for Humans or Machines.

With that in place we zap the kind imbroglio that just ensued.


>
> Alex.

Hogan, Aidan

unread,
Nov 4, 2009, 1:09:27 PM11/4/09
to pedant...@googlegroups.com
Hi,

To wrap up... We're well off-topic.

> > Hope this clarifies,
> >
> Sorry, but I didn't need a lecture. My issue is more to do with
pedantic
> dog-fooding.

My intention was not to "lecture" anyone (at least in the negative sense
of the word); my intention was to inform. I have not met you, I do not
know you or your level of expertise, and my email was trying to be clear
and concise. You seem to have interpreted my first email as negative or
patronising; that was not my intention and I am sorry that you
interpreted it as such. I was trying to directly address comments you
had made. I will try to summarise here:

"...no where is the definition is Transitivity or Symmetry inferred or


implied as part of the definition of this property. "

* the semantics of owl:sameAs and other OWL built-ins are well defined
outside of the "owl.owl" doc (as pointed out by Rinke)

"I use owl:sameAs subjectively when reasoning on it. And that's all that
people can do right now. My rule of thumb is not to assert owl:sameAs
links between resources of different types, or types that are explicitly
asserted to be disjoint."

* people should only use owl:sameAs as its semantics are defined
normatively, and not "subjectively" as I understand you mean here. If
they use owl:sameAs, they should be aware of, expect, and accept the
well-defined consequences.

Finally, the reason owl:sameAs is not defined as owl:TransitiveProperty
and owl:SymmetricProperty in "owl.owl" was due to the following design
decision from [1]:

'There is no use of OWL terms for "defining" other OWL terms, such as
"owl:complementOf".'

Although I somewhat understand their decision and could try feebly to
argue on their behalf, I won't: further discussion and justification can
be had with the people involved in [1,2].

Cheers,
Aidan

[1] http://www.w3.org/2007/OWL/wiki/Owl2DotOwlDevel
[2] http://lists.w3.org/Archives/Public/public-owl-wg/2009Jul/0029.html

Kingsley Idehen

unread,
Nov 4, 2009, 2:33:10 PM11/4/09
to pedant...@googlegroups.com
Hogan, Aidan wrote:
> Hi,
>
> To wrap up... We're well off-topic.
>
>
>>> Hope this clarifies,
>>>
>>>
>> Sorry, but I didn't need a lecture. My issue is more to do with
>>
> pedantic
>
>> dog-fooding.
>>
>
> My intention was not to "lecture" anyone (at least in the negative sense
> of the word); my intention was to inform. I have not met you, I do not
> know you or your level of expertise, and my email was trying to be clear
> and concise. You seem to have interpreted my first email as negative or
> patronising; that was not my intention and I am sorry that you
> interpreted it as such. I was trying to directly address comments you
> had made. I will try to summarise here:
>
But there is a little more context to who I am. Most important of all, I
have a litany of live owl:sameAs usage scenarios links on the Web. At
the very least, they should speak clearly about my world view and
comments. Ironically, my demos always show the use of owl:sameAs based
on the spec definitions. They also show the subjective halo (context
rules) within which this axioms inherently reside.

So step back and put yourself in my shoes. I make a post, I deliberately
don't add any links, and the immediate push back: Kingsley, do you have
something to sell? But, by not putting out a link, you get a very warped
view of the broader context etc. i.e., go lecture this guy on the
subject: owl:sameAs.

> "...no where is the definition is Transitivity or Symmetry inferred or
> implied as part of the definition of this property. "
>
> * the semantics of owl:sameAs and other OWL built-ins are well defined
> outside of the "owl.owl" doc (as pointed out by Rinke)
>
> "I use owl:sameAs subjectively when reasoning on it. And that's all that
> people can do right now. My rule of thumb is not to assert owl:sameAs
> links between resources of different types, or types that are explicitly
> asserted to be disjoint."
>
> * people should only use owl:sameAs as its semantics are defined
> normatively, and not "subjectively" as I understand you mean here. If
> they use owl:sameAs, they should be aware of, expect, and accept the
> well-defined consequences.
>

"I" meaning I make my own context rules and apply them to my own SPARQL
sessions, when I query data sources that may contain owl:sameAs
assertions. Of course, I do this knowing all consequences are mine to
bear, since the context is mine. I separate the instance data from its
data dictionary / schema / ontology all the time.

In the NYT case, if linking to a DBpedia entity of Type Person,
owl:sameAs does fine as long as NYT data space entity is also Type
Person, nice and simple. If the links are between topical entities,
DBpedia also provides nice guidance re. use of SKOS. If you want to
really deal with People as Concepts within a Concept Scheme, then look
at UMBEL [1] [2] (fwiw - OpenLink funded most of this effort).

> Finally, the reason owl:sameAs is not defined as owl:TransitiveProperty
> and owl:SymmetricProperty in "owl.owl" was due to the following design
> decision from [1]:
>
> 'There is no use of OWL terms for "defining" other OWL terms, such as
> "owl:complementOf".'
>
> Although I somewhat understand their decision and could try feebly to
> argue on their behalf, I won't: further discussion and justification can
> be had with the people involved in [1,2].
>
>

Well this whole imbroglio is a nice use case, imho.

That's said, I am also done with this matter.

Links:

1. http://umbel.org/technical_documentation.html
2. http://tr.im/E9rG -- A Description of Obama from the UMBEL world
view (i.e. treating him as a Subject Matter Concept that also happens to
be a Person). This demo has existed for eons (Linked Data time). Click
on the owl:sameAs checkbox and experience a union expansion of the
transitive kind where the underlying data corpus is 8 Billion strong
3. http://tr.im/E9pL -- for the description of a Person Subject Matter
Concept (Subject Concept for short).

Antoine Isaac

unread,
Nov 4, 2009, 6:53:04 PM11/4/09
to Rinke Hoekstra, pedant...@googlegroups.com
Hi everyone,

Quickly, a link to a related thread that has just started: http://lists.w3.org/Archives/Public/public-esw-thes/2009Nov/0000.html

Hot topic, it seems ;-)

Best,

Antoine

Axel Polleres

unread,
Nov 5, 2009, 2:51:56 AM11/5/09
to pedant...@googlegroups.com
> > My fundamental point: should we be arguing about the meaning of
> > something in prose when it has a URI that should convey its exact
> > semantics?


Aaaaaehm, didn't a guy called Goedel some decades ago already show
that a logical
language cannot fully define itself by... itself? Or did I miss
something?

So, YES, we need not only be arguing about but agree on that the
meaning of
a formal spec can define the meaning of a URI... I think that I fully
agree with Rinke on:

"URI's don't convey semantics, formal specifications do."

without the formal spec of owl: and rdfs: we can pretty much forget
about the rest of vocabulary meaning which all emerge from that
one which - quite obviously - cannot be captured in the vocabulary
itself.

Please, Kingsley, confirm that I misunderstand that someone
is doubting this axiom? But the following kinda sounds like it:

The OWL spec defines the meaning of the owl: Vocabulary. It cannot be
written in RDF, but it is also obviously not "prose":
The "prose" defining the owl: (and rdf(s): vocabulary) is actually
*mathematics*, right? ... I haven't encountered any means to express
semantics
more precise as of yet, if anybody has, then let's please dump the
term Semantic Web once and for all...

I remain confused by this discussion.

with best regards,
Axel

Kingsley Idehen

unread,
Nov 5, 2009, 7:33:48 AM11/5/09
to pedant...@googlegroups.com
Axel Polleres wrote:
>
>> > My fundamental point: should we be arguing about the meaning of
>> > something in prose when it has a URI that should convey its exact
>> > semantics?
>
>
> Aaaaaehm, didn't a guy called Goedel some decades ago already show
> that a logical
> language cannot fully define itself by... itself? Or did I miss
> something?
>
> So, YES, we need not only be arguing about but agree on that the
> meaning of
> a formal spec can define the meaning of a URI... I think that I fully
> agree with Rinke on:
>
> "URI's don't convey semantics, formal specifications do."
>
> without the formal spec of owl: and rdfs: we can pretty much forget
> about the rest of vocabulary meaning which all emerge from that
> one which - quite obviously - cannot be captured in the vocabulary
> itself.
>
> Please, Kingsley, confirm that I misunderstand that someone
> is doubting this axiom? But the following kinda sounds like it:
Axel,

I really don't know what to say anymore bar: Is there a problem with
being over specific when the desired specificity is expressible? Is
there really any harm in formally stating the obvious to the degree
possible.

The point I am failing to make is this:

1. Set context to: Newbie or Neophyte re. Semantic Web Project stuff
2. Follow a URI of a property to seek its description (to the degree it
is expressible in say, OWL, re. this matter)

The URI says, zilch, so you to to
<http://www.w3.org/TR/owl-ref/#TransitiveProperty-def>, see some excerpts:

4.4.1 owl:TransitiveProperty

When one defines a property P to be a transitive property, this means
that if a pair (x,y) is an instance of P, and the pair (y,z) is also
instance of P, then we can infer the the pair (x,z) is also an instance
of P.

Syntactically, a property is defined as being transitive by making it an
instance of the the built-in OWL class owl:TransitiveProperty, which is
defined as a subclass of owl:ObjectProperty.
....

5.2.1 owl:sameAs

The built-in OWL property owl:sameAs links an individual to an
individual. Such an owl:sameAs statement indicates that two URI
references actually refer to the same thing: the individuals have the
same "identity".

For individuals such as "people" this notion is relatively easy to
understand. For example, we could state that the following two URI
references actually refer to the same person:

<rdf:Description rdf:about="#William_Jefferson_Clinton">
<owl:sameAs rdf:resource="#BillClinton"/>
</rdf:Description>
The owl:sameAs statements are often used in defining mappings between
ontologies. It is unrealistic to assume everybody will use the same name
to refer to individuals. That would require some grand design, which is
contrary to the spirit of the web.

In OWL Full, where a class can be treated as instances of (meta)classes,
we can use the owl:sameAs construct to define class equality, thus
indicating that two concepts have the same intensional meaning. An example:

<owl:Class rdf:ID="FootballTeam">
<owl:sameAs rdf:resource="http://sports.org/US#SoccerTeam"/>
</owl:Class>
One could imagine this axiom to be part of a European sports ontology.
The two classes are treated here as individuals, in this case as
instances of the class owl:Class. This allows us to state that the class
FootballTeam in some European sports ontology denotes the same concept
as the class SoccerTeam in some American sports ontology. Note the
difference with the statement:

<footballTeam owl:equivalentClass us:soccerTeam />
which states that the two classes have the same class extension, but are
not (necessarily) the same concepts.

Then maybe you stumble across this:
Why does this exist:
<http://linkeddata.uriburner.com/about/html/http/www.w3.org/2004/02/skos/core%01narrowerTransitive
> ? I mean the inclusion of Symmetry and Transitivity in the definition?


Why can't owl:sameAs have the same clarity?

>
> The OWL spec defines the meaning of the owl: Vocabulary. It cannot be
> written in RDF, but it is also obviously not "prose":
> The "prose" defining the owl: (and rdf(s): vocabulary) is actually
> *mathematics*, right? ... I haven't encountered any means to express
> semantics
> more precise as of yet, if anybody has, then let's please dump the
> term Semantic Web once and for all...
>
> I remain confused by this discussion.

I really think we should end this discussion at this juncture. This
whole matter has simply taken me to place I simply don't need to be at
right now.

Kingsley

Axel Polleres

unread,
Nov 5, 2009, 10:53:52 AM11/5/09
to pedant...@googlegroups.com
Dear Kingsley,

On 5 Nov 2009, at 04:33, Kingsley Idehen wrote:

> Why can't owl:sameAs have the same clarity?


Just allow me a last remark on that, I agree we don't want to
overstress this discussion.
it's basically, because http://www.w3.org/TR/owl-ref/
doesn't define the semantics of OWL, but

http://www.w3.org/TR/owl-semantics/

does. I guess we are in agreement that standard specs don't read
nicely, but at least that
one is clear from the viewpoint of what it formally means.

best regards,
Axel

Kingsley Idehen

unread,
Nov 5, 2009, 11:12:03 AM11/5/09
to pedant...@googlegroups.com
Axel,

Yes, I guess my point may have been a little mangled (I really assumed
juxtaposition of skos:narrowerTransitive and owl:sameAs URIs would
accentuate the point).

I was trying to say is post: de-referencing the property URI, and
googling on "owl:sameAs" getting a doc, reading it, its *possible* and
*typical* to conclude: owl:sameAs is at best an assertion you can make
subjective as the data space owner. You can take the view that Reasoning
is subjective.

Now if you toss out all of the above, and my initial comments. How do my
comments imply "ignorance of owl:sameAs" when I have a product that
actually implement "owl:sameAs" as all the lectures I received espouse.
Not only does it do that, I put out a litany of links that actually
demonstrate "owl:sameAs" virtues (but within context rules boundaries).
Even worse, and most ironic, I painstakingly worked through many
iterations of this feature in Virtuoso over the years, culminating in
simple UIs for demonstrating what "owl:sameAs" offers.

Context can be fluid, but I really think my true intention has long been
lost in this conversation.

I am sure you know, or least I hope, my biggest concern is the newcomer
and skeptics audience. Thus, I just want us to dog-food a little better.

Also, and not to hammer this point, if I truly believe that owl:sameAs
isn't Transitive or Symmetrical, and I sure Google would provide ample
evidence from past commentary, that's one of the beauties of the Web.
These comments are recent, and with fixing the URI and docs in mind.

Axel Polleres

unread,
Nov 5, 2009, 11:16:35 AM11/5/09
to pedant...@googlegroups.com
p.s. ... if something was learnt from this discussion... and in order
to be *very* pedantic.
we could ask the OWL WG to add rdfs:isDefinedBy links to the semantics
document?

i.e. instead of

<rdf:Property rdf:about="&owl;sameAs">
<comment>The property that determines that two given
individuals are equal.</comment>
<domain rdf:resource="&owl;Thing"/>
<isDefinedBy rdf:resource="&owl;"/>
<label>sameAs</label>
<range rdf:resource="&owl;Thing"/>
</rdf:Property>

<rdf:Property rdf:about="&owl;sameAs">
<comment>The property that determines that two given
individuals are equal.</comment>
<domain rdf:resource="&owl;Thing"/>
<isDefinedBy rdf:resource="http://www.w3.org/TR/owl2-rdf-based-semantics/
"/>
<label>sameAs</label>
<range rdf:resource="&owl;Thing"/>
</rdf:Property>

Would that make sense? (not sure, they actually do have)

<rdfs:isDefinedBy rdf:resource="http://www.w3.org/TR/owl2-rdf-based-semantics/
" />

in the ontology header of
http://www.w3.org/2002/07/owl#

which is probably enough...

Axel

Reply all
Reply to author
Forward
0 new messages