Re: Ontology editor + why RDF?

This is a wonderful discussion, and very apropos for this list. As we  
think about "semantic web use cases" for HCLS, we find that our  
imaginations need just this kind of stretching to conceive of good  
applications -- not because applications don't exist (in spades) --  
but because we're just discovering them, and so finding them means  
breaking old thought habits.

There's an important point that comes out of this about how to  
discuss SW with "nonbelievers". Some ask "What is it that the SW can  
do, that I can't do with ..." (fill in the blank: database merging,  
data modeling standards, schema languages, blah blah).

(I happen to be out at SMI at a Protege course at the moment and  
Jennifer Vendetti made exactly this point about ontologies, so I'm  
shamelessly stealing her point...)

One true answer to this question is "nothing". Another true answer is  
"many, many things!!"

"Nothing" ---> RDF/RDFS/OWL are formally equivalent to a subset of  
what can be expressed in relational database language. RDBMS are in  
this sense "more powerful" than any "semantic web language". (I think  
this is what Vipul meant when he said that moving knowledge out onto  
the web does not "add any semantics".)

"Many,, many things":  RDF/RDFS/OWL make *your* assertions accessible  
to hundreds of millions of people, who can add their own assertions.  
Millions of new assertions about your resource --- even if any  
individual one is not incremementally very information-rich -- can  
collectively be worth a lot more than a mere handful of "strong"  
assertions that only a few people can "see". (And if the former  
requires minimal effort on any one person's part, and the latter  
requires herculean effort on one person's part, then the former is a  
lot more likely to happen than the latter.) (This is (I think) what  
Jim means by a little goes a long way.)

The point is, the original question is phrased in an unhelpful way.  
The real question is not "what can it do", but "does it make  
interoperability easier, and hence the creation of useful semantic  
communities more likely". And these new semantic communities will  
change the way we do health science, and the way we live -- and hence  
how they will look is hard to conceive of in advance !!!!

John


On Mar 31, 2006, at 1133, John Barkley wrote:

>
>
> I believe there is an additional advantage beyond semantics and the  
> "web".
>
> I would suggest that the quality of information models can be  
> significantly
> improved using semantic web methods and tools as compared to those  
> developed
> using other methods in common use, e.g., relational, XML, UML.  With
> semantic web methods and tools, more of an information model can be  
> made
> explicit in a formal language while the model remains amenable to  
> automated
> validation and testing. In other words, things about an information  
> model
> that can only said implicitly with other methods (e.g., in  
> documentation)
> can be said explicitly with semantic web methods while still  
> maintaining the
> capability for applying fully automated validation and testing to  
> the model.
>
>
>
> The reason for this is the work that has been done in the theory of
> Description Logic on which the semantic web rests. Description logic
> provides the foundation on which formal languages for describing  
> information
> models may be defined. RDF/OWL is the standard description logic  
> formal
> language of the semantic web.
>
>
>
> Thus, as you know, a knowledge base described in RDF/OWL has the  
> following
> characteristics:
>
>
>
> 1.      Whether the knowledge base is amenable to automated  
> validation and
> testing can be readily determined by a grammatical examination of the
> features of RDF/OWL used in the description. OWL DL is the  
> sublanguage of
> RDF/OWL which supports automated validation and testing.
>
>
>
> 2.      Given that a knowledge base is amenable to automated  
> validation and
> testing, a fully automated "reasoner"  can used to perform such  
> validation
> and testing.
>
>
> Specification of an information model in a formal language with the
> certainty that the validity of such a specification can be tested
> automatically can go a long way in improving the quality of the model.
>
> jb
>
>
> ----- Original Message -----
> From: "Kashyap, Vipul" <VKASHYAP1@PARTNERS.ORG>
> To: "Danny Ayers" <danny.ayers@gmail.com>
> Cc: "Jim Hendler" <hendler@cs.umd.edu>; "deWaard, Anita (ELS)"
> <A.dewaard@elsevier.com>; <public-semweb-lifesci@w3.org>
> Sent: Friday, March 31, 2006 8:08 AM
> Subject: RE: Ontology editor + why RDF?
>
>
>
>> I saw a quote not long ago, not sure of the source (recognise this
>> Jim?), approximately: "what's new about the Semantic Web isn't the
>> semantics but the web".
>
> [VK] This is a great quote and expresses clearly that the value  
> proposition
>      in representing and linking vocabularies using URIs stems from  
> the
>      Web more than "semantics"
>
>> I take VK's point that this in itself isn't going to convince many IT
>> folks. I think the big persuader there is data integration, even on a
>> sub-enterprise kind of scale.
>
> [VK] Agreed, one of the clearer value propositions is data  
> integration.
>
>> Being able to use ontologies to infer new information is a massive
>> plus (I imagine especially in the lifesciences). Bigger still are the
>> (anticipated) benefits of the Semantic Web when the network effect
>> kicks in. But the ability to use RDF to simply merge data from
>> multiple sources consistently (and query across it), without needing
>> complete up-front schema design is a very immediate, tangible gain.
>>
>> The work done around SKOS (and specific tasks like expressing WordNet
>> in RDF) does suggest RDF/OWL is a particularly good technology choice
>> for thesauri.
>
> [VK] Danny, has articulated some potential benefits:
>      - Network effects
>      - Schema-less linking based data integration
>
> I would argue that both these benefits stem from the web  
> infrastructure and
> have
> nothing to do with the "semantics" of anything per-se.
>
> Also, one could argue that having a standardized markup language  
> whether it
> be even HTML or XML enables the above to a significant extent.
>
> So the value proposition question could be:
>
> What is it about RDF that enables network effects and schema less data
> linking
> better than HTML, relational tables or XML in a more significant  
> manner?
>
> Is the improvement enabled v/s the cost required to achieve it an  
> attractive
> trade off?
>
> Look forward to yours and the groups responses to these questions.
>
> Cheers,
>
> ---Vipul
>
>
>

Received on Friday, 31 March 2006 22:08:42 UTC