Using openguid to link non-formal descriptions
bernard
described by formal descriptions in RDF vocabularies. I wonder if it
would not be a good idea to extend its use to any kind of resources,
including resources without declared formal semantics. Such an
extension would maybe help the community to understand the (absence of
formal) semantics of such links, and clarify the confusion seen in
some questions and criticisms that have been thrown already.
I have discovered and tried today the amazing service TinEye at
http://tineye.com, which finds on the Web copies and avatars of an
image defined by its URI or uploaded from your system. I was striken
by the similarity of service with Open GUID. Copies and avatars of an
image have globally the same referent, either the original image, or
the object they are an image of. So URIs of all those images could
actually be linked through oguid:identical links. Actually TinEye
provides opaque URI for any search, which are quite similar to GUID
URIs
This can be applied to any resource that is clearly dedicated to be
the representation/description of one thing. Think about people or
organizations profiles, product descriptions, pages describing a
place, species description.
Even if there is no available formal description of those resources,
seems to me correct to assert things like
http://www.linkedin.com/companies/64279/Mondeca
oguid:identical http://www.mondeca.com/
http://www.eol.org/taxa/17183620 oguid:identical
http://mark.gattone.free.fr/champignons/pr%E9sentation/fiches/amanite%20vireuse.jpg
The latter example seems to me particularly interesting, since it
links a structured, almost formal description of a species, with an
image representing the same species. We have there completely
different ways to point at a referent, but the referent is clearly the
same.
Thoughts?
Bernard
ja...@openguid.net
I watched their video, and it's cool how they can find photoshopped
versions of the image, which of course one could argue is no longer
the same image.
But your points remains valid, associating various web resources is
compelling. With the disclaimer that the resources must be clearly
dedicated to the representation of one thing, it seems like the same
logic behind identical property.
I have no idea the impact of using RDF to link non RDF resources, so
maybe some other construct is needed. And I'm not sure the Open GUID
Services would be a good place to store such links, but it's
definitely something to think about.
To date, my thinking was the RDFa "about" attributes in html documents
are a natural place to link Open GUIDs. Without double checking, I
think they can be placed on img tags as well? Of course that requires
some leg work by content authors, which is no easy request (even with
nice tools). There probably is a case for some method to add those
links outside the resources.
Thanks for the idea.
On Oct 3, 3:48 pm, bernard <bernard.vat...@mondeca.com> wrote:
> So far oguid:identical has been suggested to link only resources
> described by formal descriptions in RDF vocabularies. I wonder if it
> would not be a good idea to extend its use to any kind of resources,
> including resources without declared formal semantics. Such an
> extension would maybe help the community to understand the (absence of
> formal) semantics of such links, and clarify the confusion seen in
> some questions and criticisms that have been thrown already.
> by the similarity of service with Open GUID. Copies and avatars of an
> image have globally the same referent, either the original image, or
> the object they are an image of. So URIs of all those images could
> actually be linked through oguid:identical links. Actually TinEye
> provides opaque URI for any search, which are quite similar to GUID
> URIs
> This can be applied to any resource that is clearly dedicated to be
> the representation/description of one thing. Think about people or
> organizations profiles, product descriptions, pages describing a
> place, species description.
> Even if there is no available formal description of those resources,
> seems to me correct to assert things like
>
> http://www.linkedin.com/companies/64279/Mondeca
Ryan Shaw
> So far oguid:identical has been suggested to link only resources
> described by formal descriptions in RDF vocabularies. I wonder if it
> would not be a good idea to extend its use to any kind of resources,
> including resources without declared formal semantics.
This would be a good way to build bridges to things like Freebase
topics and WorldCat Identities. I have been doing this using various
adhoc properties; it would be nice to have something more widely used.
Bernard Vatant
> Interesting. Tineye looks very slick.
>
> I watched their video, and it's cool how they can find photoshopped
> versions of the image, which of course one could argue is no longer
> the same image.
>
image or not. As resources, versions should certainly not be considered
the same, most of the time. But if you consider the referent of images
of a painting in a museum, it's hard to sort copies of the same original
shot from copies of different shots. But it does not really matte, the
original painting being the same is certainly the most interesting point
in this case, since the painting acts as the common referent. Is there
really a difference with copies of the same original numeric image? I
don't think so. They have also a common referent, the original image.
> But your points remains valid, associating various web resources is
> compelling. With the disclaimer that the resources must be clearly
> dedicated to the representation of one thing, it seems like the same
> logic behind identical property.
>
of such pages, and they could be either tagged as such by their
publisher (using GUID or any kind of non-ambiguous tagging), or even
automatically discovered by smart search engines.
> I have no idea the impact of using RDF to link non RDF resources, so
> maybe some other construct is needed.
soon as a resource has a URI, otherwise said, as soon as it's identified
as a resource, it can be in RDF the subject or object of a triple.
> And I'm not sure the Open GUID
> Services would be a good place to store such links,
> but it's
> definitely something to think about.
>
> To date, my thinking was the RDFa "about" attributes in html documents
> are a natural place to link Open GUIDs. Without double checking, I
> think they can be placed on img tags as well?
> Of course that requires
> some leg work by content authors, which is no easy request (even with
> nice tools). There probably is a case for some method to add those
> links outside the resources.
>
Bernard
--
*Bernard Vatant
*Senior Consultant
Vocabulary & Data Engineering
Tel: +33 (0) 971 488 459
Mail: bernard...@mondeca.com <mailto:bernard...@mondeca.com>
----------------------------------------------------
*Mondeca**
*3, cité Nollez 75018 Paris France
Web: www.mondeca.com <http://www.mondeca.com>
Blog: Leçons de Choses <http://mondeca.wordpress.com/>
----------------------------------------------------**
Simon Reinhardt
> > I watched their video, and it's cool how they can find photoshopped
> > versions of the image, which of course one could argue is no longer
> > the same image.
>
> Yes, indeed, one can argue at length on what it means to be the same
> image or not. As resources, versions should certainly not be considered
> the same, most of the time. But if you consider the referent of images
> of a painting in a museum, it's hard to sort copies of the same original
> shot from copies of different shots. But it does not really matte, the
> original painting being the same is certainly the most interesting point
> in this case, since the painting acts as the common referent. Is there
> really a difference with copies of the same original numeric image? I
> don't think so. They have also a common referent, the original image.
or that their referent is identical and to me a property that doesn't
clearly say which of those two it is about is as useless as no
linkage. It basically comes down to the "thing vs. document describing
the thing" problem which I thought was solved in the Semantic Web
community by now. So will oguid:identical smudge this again by being
about... both? I guess I don't really understand the definition of the
property. From the spec:
"The property oguid:identical indicates two RDF resources refer to the
same thing."
As you pointed out there are no RDF resources. Any resource *is* a
thing. But not every resource refers to something else. What does the
resource denoted by the URI <http://www.w3.org/People/Berners-Lee/
card#i> refer to? I can't think of anything that Tim would refer to,
he's just a person. So you would have to restrict domain and range of
the property to resources that refer to others. But then you can't use
it anymore to state that two things are identical *themselves* (the
resource denoted by the URI <http://www4.wiwiss.fu-berlin.de/
bookmashup/persons/Tim+Berners-Lee> *is* the resource denoted by the
URI <http://www.w3.org/People/Berners-Lee/card#i>).
Then you also have different levels of talking about the images
themselves, but that's also mostly solved with the translation of FRBR
into RDF: http://purl.org/vocab/frbr/core
This shouldn't be mixed up either. So while it would sure be nice to
have a generic property that isn't at risk of being too restrictive by
getting into detail too much and talking too much about border cases,
I'm not sure about the actual machine-understandable value of it if it
isn't clear about which level of "identical" it's talking about or
what thing it actually is that is considered identical.
Simon
Jason Borro
> There's a huge difference between saying that two images are identical
> or that their referent is identical and to me a property that doesn't
> clearly say which of those two it is about is as useless as no
> linkage. It basically comes down to the "thing vs. document describing
> the thing" problem which I thought was solved in the Semantic Web
> community by now. So will oguid:identical smudge this again by being
> about... both? I guess I don't really understand the definition of the
> property. From the spec:
> "The property oguid:identical indicates two RDF resources refer to the
> same thing."
> As you pointed out there are no RDF resources. Any resource *is* a
> thing. But not every resource refers to something else. What does the
> resource denoted by the URI <http://www.w3.org/People/Berners-Lee/
> card#i> refer to? I can't think of anything that Tim would refer to,
> he's just a person. So you would have to restrict domain and range of
> the property to resources that refer to others. But then you can't use
> it anymore to state that two things are identical *themselves* (the
> resource denoted by the URI <http://www4.wiwiss.fu-berlin.de/
> bookmashup/persons/Tim+Berners-Lee> *is* the resource denoted by the
> URI <http://www.w3.org/People/Berners-Lee/card#i>).
document" confusion (whether it's clear enough in the spec is another
matter...possibly not since you are asking).
To be clear, it is the "thing" that is declared identical. As I see it,
all resources on the web are representations (be it textual, an image,
or a formal OWL definition). In other words, Tim does not exist as a
resource, only as a person; even the first URI you list is not him, but
a description of him. oguid:identical simply says that these resources
are all descriptions of the same real-world concept. An Open GUID for
Tim also refers to him, but carries no formal semantics...it is the
simplest definition possible needed to distinguish him from other people
and organic food.
Bernard was pointing out that the phrase "RDF resources" was too
restrictive, because it's not just RDF representations that could be
declared identical (as in referring to the same thing). Maybe the
property should be named "sameReferent"? I chose identical merely
because it fit with the theme of a global identifier.
>
> Then you also have different levels of talking about the images
> themselves, but that's also mostly solved with the translation of FRBR
> into RDF: http://purl.org/vocab/frbr/core
> This shouldn't be mixed up either. So while it would sure be nice to
> have a generic property that isn't at risk of being too restrictive by
> getting into detail too much and talking too much about border cases,
> I'm not sure about the actual machine-understandable value of it if it
> isn't clear about which level of "identical" it's talking about or
> what thing it actually is that is considered identical.
Using oguid:identical for for images would imply that the images resolve
in people's heads to the same concept. It wouldn't say anything about
the particular image file, e.g.
In the case of a photoshopped mona lisa, people will still recognize the
original referent, but it has arguably taken on a different identity as
"a funny version of the mona lisa". In this case, the real-world
concept is "a funny version of the mona lisa"; it is still "real" in
that people can think about it. The URI that points to the file
represents the new concept as a set of pixels. It is definitely a
fringe case that might not have much value in the Linked Data world.
> Simon
> >
>
>
Simon Reinhardt
> To be clear, it is the "thing" that is declared identical. As I see it,
> all resources on the web are representations (be it textual, an image,
> or a formal OWL definition). In other words, Tim does not exist as a
> resource, only as a person; even the first URI you list is not him, but
> a description of him. oguid:identical simply says that these resources
> are all descriptions of the same real-world concept. An Open GUID for
> Tim also refers to him, but carries no formal semantics...it is the
> simplest definition possible needed to distinguish him from other people
> and organic food.
To me the URI <http://www.w3.org/People/Berners-Lee/card#i> doesn't
point to a description of Tim (that would be <http://www.w3.org/People/
Berners-Lee/card>, the document (information resource) that describes
the former) but it points to Tim himself. So in the semiotic triangle
(<http://en.wikipedia.org/wiki/Triangle_of_reference>) it would be the
referent (unless you explicitly state otherwise). I guess to you it
would be the concept of / reference to / idea of Tim. In my eyes, the
document *describes* a certain idea of Tim from a certain point of
view, but there still is no URI for this view, just for a document
which contains descriptions of the view.
Also I wouldn't say that <http://www.w3.org/People/Berners-Lee/card#i>
points to a resource on the Web, it is just an identifier that can
lead you to a resource on the Web through several steps.
Simon
Bernard Vatant
I thinl we are at the heart of the issue here, and I answer here also to
your question in the thread "identical vs sameAs"
> On Oct 6, 4:47 am, Jason Borro <ja...@openguid.net> wrote:
>
>> To be clear, it is the "thing" that is declared identical. As I see it,
>> all resources on the web are representations (be it textual, an image,
>> or a formal OWL definition).
>>
Indeed! All resources are representations! A text is a documentation, a
photograph is a representation, a RDF description is a representation.
To put it more exactly in semiotic terms, the URI - resource pair is a
sign : The URI is the signifier, and the resource is the signified. See
below ...
>> In other words, Tim does not exist as a
>> resource, only as a person; even the first URI you list is not him, but
>> a description of him. oguid:identical simply says that these resources
>> are all descriptions of the same real-world concept. An Open GUID for
>> Tim also refers to him, but carries no formal semantics...it is the
>> simplest definition possible needed to distinguish him from other people
>> and organic food.
>>
>
> This is where we have quite different views of the world then. :-)
> To me the URI <http://www.w3.org/People/Berners-Lee/card#i> doesn't
> point to a description of Tim (that would be <http://www.w3.org/People/
> Berners-Lee/card>, the document (information resource) that describes
> the former) but it points to Tim himself.
Indeed you say it yourself : it *points* to Tim, it is not Tim. Even if
it identifies Tim, it *is* not Tim. Take it anyway you want, there is
always a level of indirection.
You are not "closer" to Tim because you provided a formal identifier and
description. You're always dealing with signs. On the Web, like in any
language or information system, you deal with signs. The referents are
outside the URI and resources system are they are outside the language.
The best you can do is to say : although the referents are outside the
representation system, we can agree that those signs have the same
referent ... even if in our system they belong to different grammars of
representation, e.g. a formal description vs a photograph.
> So in the semiotic triangle
> (<http://en.wikipedia.org/wiki/Triangle_of_reference>) it would be the
> referent (unless you explicitly state otherwise).
The referent is outside the system. There is no semantics on referents,
only on signs.
> I guess to you it
> would be the concept of / reference to / idea of Tim. In my eyes, the
> document *describes* a certain idea of Tim from a certain point of
> view, but there still is no URI for this view, just for a document
> which contains descriptions of the view.
> Also I wouldn't say that <http://www.w3.org/People/Berners-Lee/card#i>
> points to a resource on the Web, it is just an identifier that can
> lead you to a resource on the Web through several steps.
>
My view (and Jason's) is that this is true of *any* URI. Unless you
think you captured the referent in the system ...
Simon Reinhardt
> On Oct 6, 4:47 am, Jason Borro <ja...@openguid.net> wrote:
> >> To be clear, it is the "thing" that is declared identical. As I see it,
> >> all resources on the web are representations (be it textual, an image,
> >> or a formal OWL definition).
>
> Indeed! All resources are representations! A text is a documentation, a
> photograph is a representation, a RDF description is a representation.
> To put it more exactly in semiotic terms, the URI - resource pair is a
> sign : The URI is the signifier, and the resource is the signified. See
> below ...
resources being representations. A resource may have a representation
or not (an image of it, a description of it). Another resource (the
image, the RDF document) may be a representation of a resource. But
clearly, resources are not restricted to that, that's the whole point
of the Semantic Web: to talk about things (using URIs) which are not
necessarily representations of other things.
> > This is where we have quite different views of the world then. :-)
> > To me the URI <http://www.w3.org/People/Berners-Lee/card#i> doesn't
> > point to a description of Tim (that would be <http://www.w3.org/People/
> > Berners-Lee/card>, the document (information resource) that describes
> > the former) but it points to Tim himself.
>
> Indeed you say it yourself : it *points* to Tim, it is not Tim. Even if
> it identifies Tim, it *is* not Tim. Take it anyway you want, there is
> always a level of indirection.
Tim, I didn't say the resource points to him. You're mixing up the URI
and the resource it identifies here!
Of course the URI isn't Tim. It is, as you said, a signifier. It's a
random string, it doesn't matter at all. In RDF the URIs never have
any semantics attached to them, you don't talk about them, you only
ever talk about the resources they point to. The URIs are just a means
to point to resources and tell them apart.
> You are not "closer" to Tim because you provided a formal identifier and
> description. You're always dealing with signs. On the Web, like in any
> language or information system, you deal with signs. The referents are
> outside the URI and resources system are they are outside the language.
> The best you can do is to say : although the referents are outside the
> representation system, we can agree that those signs have the same
> referent ... even if in our system they belong to different grammars of
> representation, e.g. a formal description vs a photograph.
information system, only the URI is. An image of him would just be
another way of pointing to him. The image points by description, the
URI points by reference. It is however important to keep those two
ways of pointing at things apart, i.e. you have to explicitly state
which of them you are using. If you want to use an image to point to
the thing that can be seen on it then you have to say that, otherwise
you're talking about the image, not its subject.
> > So in the semiotic triangle
> > (<http://en.wikipedia.org/wiki/Triangle_of_reference>) it would be the
> > referent (unless you explicitly state otherwise).
>
> The referent is outside the system. There is no semantics on referents,
> only on signs.
I get the feeling that we just use a different language but still talk
about the same things. So it's hard to tell what we actually disagree
on. :-)
Simon
ja...@openguid.net
> information system, only the URI is. An image of him would just be
> another way of pointing to him. The image points by description, the
> URI points by reference. It is however important to keep those two
> ways of pointing at things apart, i.e. you have to explicitly state
> which of them you are using. If you want to use an image to point to
> the thing that can be seen on it then you have to say that, otherwise
> you're talking about the image, not its subject.
property explicitly denotes equality in the subject. The subject of a
resource identified by a URI, be it an image or RDF document. Tim is
not the resource, but the subject of that resource.
Ideas on making this clearer in the spec would be most helpful.
Heiko Stoermer
Heiko Stoermer
University of Trento, Italy
Dept. of Information Science and Engineering (DISI)
http://disi.unitn.it/~stoermer
OKKAM id:
http://www.okkam.org/entity/ok5f23a5ce-a683-4c4d-ae73-b78cdc17aec1
Bernard Vatant
As Jason pointed somewhere in a previous message, oguid:identical could
be extended to any kind of resource(s), *provided such resources are
explicitly designed by their publisher to be the representation of one
single thing*. Such a definition exclude your examples of photographs of
several people of which one happens to be Tim (unless they are used to
represent a team). The resources likely to be usefully and
non-ambiguously linked by oguid:identical must have been explicitly
designed to represent one thing. That's what of course resources used in
RDF are for, but come also to mind :
- People, companies, projects profiles / home pages
- Resources describing a specific product/brand/trademark, or a specific
item
- Resources describing a specific place, or spatially located object or
service : an hotel, a room in a museum, a railway station
- Resources describing a specific living species, either by a structured
description, either by picture. See my mushroom example in a previous
message.
Of course such resources can always be interpreted to refer to other
things, like a t-shirt. But this is true also for a RDF description. If
I include a triple in TBL's description including one of his
publication, I can say this resource is also a representation of this
very publication. Just a question of viewpoint on the resource.
And of course Wikipedia pages was the best available set of such
resources, since, by design, a Wikipedia page is about a single thing.
Wiki pages are subject-centric. So it was easy (conceptually at least,
if not techically) to build DBpedia URIs, based on this very "single
about-ness".
The identity of referent has to be the result of a social agreement.
IOW oguid:identical doesn't assert identity between an image and a
thing, or a concept and a class ... it does not assert identity of signs
(the resources), it attests identity of referents of resources which
have been designed as *signs* of one thing.
Well ... yet another attempt to explain that. :-\
Heiko Stoermer a écrit :
--
*Bernard Vatant