W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > March 2013

Re: owl:sameAs - Is it used in a right way?

From: John Madden <john.madden@duke.edu>
Date: Sat, 16 Mar 2013 12:59:10 -0400
Cc: David Booth <david@dbooth.org>, Jeremy J Carroll <jjc@syapse.com>, Kingsley Idehen <kidehen@openlinksw.com>, w3c semweb HCLS <public-semweb-lifesci@w3.org>, Jim McCusker <mccusj@rpi.edu>
Message-Id: <8E30947A-6CDA-4285-BC6E-A8D0AFF696CD@duke.edu>
To: Umutcan ŞİMŞEK <s.umutcan@gmail.com>
Medical records are filled with modal assertions:

	Possibly P(x)
	I believe that P(x)
	Jim believes P(x) (whereas e..g. perhaps David, Umutcan, Jeremy and I don't).
	At 5:00 pm today P(x)
	I disavow P(x)
	It is extremely unlikely that P(x)
	I know that P(x)
	I regret that P(x)
	I am reponsible for bringing about P(x) (!!!!)

and these are typically the most "interesting" (in the sense of having practical medical consequences) statements in medical records. 

OWL logic (out-of-the-box) does not include support for modal reasoning of arbitrary kind (and there are a countless number of kinds). Put another way, OWL out-of-the-box has a commitment to one particular modality. So we don't expect an OWL reasoner to reason with arbitrary modal assertions; in fact, we don't expect the OWL language necessarily to even be competent to express multiple, arbitrary modalities, out-of-the box. It just isn't part of the native language.

I agree with Jim and others that if you want to use OWL, you must let OWL be OWL. We should reason locally with it, accepting its limitations. One such limitation is that if you choose to reason with owl:sameAs, under OWL rules (i.e under OWL modality), you have situated yourself within a universe, consisting of a set of possible worlds related to each other in a particular way (to cast it that way), in which the resources referenced really are the same resource in all relevant respects—where "relevant" means relevant to your considerations: considerations that are not part of OWL, but of which OWL inference rules are perforce a subset.

If you are in doubt whether you can buy into that, then you just shouldn't include those particular triples—or else not use OWL (out-of-the box, or perhaps at all) to reason. (Maybe "reason" some other way, maybe "manually" by using your noggin while inspecting the triples or some insight-provoking representation thereof.) Or better and in addition, you should simply consider the result of any OWL reasoning exercise as a kind of experiment—not "truth" simpliciter, but just a way of informing yourself about the implications of situating yourself within some set of possible worlds under OWL modality, given that you provisionally accept certain assertions as facts.

In my long-held opinion, where clinical records are concerned "local" and "(OWL-)relevant" would often mean pre-selecting a pretty darned small set of "wild-type" triples, by which I mean triples culled from sundry sources in the jungle of the Semantic Web: a few dozen? a few hundred? Maybe. Maybe more or less, depending on what it is you hope to accomplish when you press the fateful button labeled "INFER".

Of course, it's possible to fiat-define as many modal predicates as you want, and to use them to navigate through the jungle; but not to automagically reason with them. Fiat predicates like <asserts> (with domain e.g. foaf:Person and range e.g. trix:graph; thank you Jeremy) could very useful for pre-navigating among graph fragments to select the ones with which you care to populate your particular world(s).

John

On Mar 16, 2013, at 1:08 AM, Jim McCusker <mccusj@rpi.edu> wrote:

> David,
> 
> The problem with this is that by definition, URIs ALWAYS denote the same resource. If there is doubt that you might be denoting something other than what a resource is, you should be defining your own resource.
> 
> Jim
> 
> 
> On Sat, Mar 16, 2013 at 12:35 AM, David Booth <david@dbooth.org> wrote:
> Hi Umutcan,
> 
> You have indeed stumbled on a deep question, and I think Jeremy's suggestion is exactly right.  This paper on "Resource Identity and Semantic Extensions:
> Making Sense of Ambiguity" illustrates how owl:sameAs works in RDF semantics:
> http://dbooth.org/2010/ambiguity/paper.html#sameAs
> 
> There are two keys to understanding owl:sameAs.  One is to answer the question: what RDF graph are you considering?  The other is to understand that the same URI may denote different things in different RDF graphs.  It is only when RDF statements are in the *same* graph that the RDF semantics requires the URI to denote the same resource.  That is why the question of what graph you are considering is crucial, and why Jeremy suggested keeping the different perspectives in different graphs.
> 
> FYI, the above paper also explains how you can "split" the identity of an RDF resource if you need to merge RDF graphs that use the same URI in contradictory ways.
> 
> David
> 
> 
> 
> On 03/15/2013 02:29 PM, Jeremy J Carroll wrote:
> I did not find this a rookie question at all.
> 
> This seems to get to the heart of some of the real difficult issues in Semantic Web.
> 
> My perspective is different from yours, and a resource description that I author is a description of the resource from my perspective; a resource description that you author is a description from your perspective.
> 
> If I have some detailed application that depends in some subtle way on my description, I may want to ignore your version; on the other hand, a third party might want to use both of our points of view.
> 
> One way of tacking this problem is to have three graphs for this case:
> 
> Gj, Gu, G=
> 
> Gj contains triples describing my point of view
> Gu contains triples describing your point of view
> G= contains the owl:sameAs triples
> 
> Then, in some application contexts, we use Gj, sometimes Gu, and sometimes all three.
> 
> Jeremy
> 
> 
> 
> 
> On Mar 15, 2013, at 11:02 AM, Umutcan ŞİMŞEK <s.umutcan@gmail.com> wrote:
> 
> Thanks for the quick answer : )
> 
> So this issue is that subjective for contexts which allows to use owl:sameAs to link resources  if they are not semantically even a little bit related in real world?
> 
> Sorry if I'm asking too basic questions. I'm still a rookie at this :D
> 
> Umutcan
> 
> 
> On 15-03-2013 19:38, Kingsley Idehen wrote:
> On 3/15/13 1:05 PM, Umutcan ŞİMŞEK wrote:
> My question is, does LODD use owl:sameAs properly? For instance, are those two resources, dbpedia:Metamizole and drugbank:DB04817 (code for Metamizole), really identical? Or am I getting the word "property" in the paper wrong?
> The question is always about: do those URIs denote the same thing? Put differently, do the two URIs have a common referent?
> 
> ## Turtle ##
> 
> <#i> owl:sameAs <#you>.
> 
> ## End ##
> 
> That's a relation in the form of a 3-tuple based statement that carries entailment consequences for a reasoner that understand the relation semantics. Through some "context lenses" the statement above could be accurate, in others totally inaccurate.
> 
> Conclusion, beauty lies eternally in the eyes of the beholder :-)
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> -- 
> Jim McCusker
> Programmer Analyst
> Krauthammer Lab, Pathology Informatics
> Yale School of Medicine
> james.mccusker@yale.edu | (203) 785-4436
> http://krauthammerlab.med.yale.edu
> 
> PhD Student
> Tetherless World Constellation
> Rensselaer Polytechnic Institute
> mccusj@cs.rpi.edu
> http://tw.rpi.edu
Received on Saturday, 16 March 2013 16:59:38 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:53:01 UTC