Re: Advancing translational research with the Semantic Web

Bijan Parsia wrote:
> 
> On May 16, 2007, at 11:15 AM, Phillip Lord wrote:
> 
>>
>>>>>>> "EJ" == Eric Jain <Eric.Jain@isb-sib.ch> writes:
>>
>>   EJ> Just catching up on reading papers :-)
>>
>>   EJ> <http://www.biomedcentral.com/1471-2105/8/S3/S2>
>>
>>   EJ> "It is also useful to know who believes something and
>>   EJ> why. However, there is no standard way of expressing such
>>   EJ> information about a statement [...]"
>>
>>   EJ> Reification?
>>
>> That's who, not why.
> 
> No, you can do both with reification.
> 
>> The Gene Ontologies evidence codes are and
>> references are much closer.
>>
>> Also, I am not sure of the semantics of reification.
> 
> RDF reification has very little to no built in semantics. What it 
> provides is a standardized syntax.
> 
>> Does it mean "I
>> made this statement", "I believe this statement" or "I am the person
>> responsible for the evidence on which this statement is based".
> 
> It doesn't provide any of these propositional attitudes, but you could 
> layer these on top of reification, i.e., use reification of a triple as 
> the object of an "I believe..." statement. You could make a little 
> "belief object" which had properties for who, when, why, to what degree, 
> etc and "content" which would be the reified triple.
> 
> You can, of course, role your own reification mechanism, and that's what 
> one typically does, e.g., to get n-ary predicates (or for things where 
> one might have used an n-ary pred). Some stores (to my surprise, e.g., 
> Jena) will take reified triples and condense them into a nicer internal 
> form for querying (obviously this wouldn't help with a role-your-own).
> 
> In either case you have to supply the semantics of the operators 
> separately, either through axioms or a special tool support.
> 
> OWL 1.1 provides a mechanism for putting annotations directly on axioms 
> (which in the RDF mapping is encoded using RDF reification), with no 
> specified semantics. One could use these as a hook, or just extend the 
> XML or functional syntax directly.
> 
> You could also use literals.
> 
> However, all this *supports* your point. There *IS* no standardized way 
> to represent this sort of information. There is a more or less standard 
> (and widely loathed) hook/technique upon which you could build a 
> standard mechanism for representing this sort of information.

It's worth making sure everyone's aware of the RDF Semantics spec text 
on this, http://www.w3.org/TR/rdf-mt/#Reif to avoid reading too much 
into the reification vocabulary. Specifically,

"""Note that this way of understanding the reification vocabulary does 
not interpret reification as a form of quotation. Rather, the 
reification describes the relationship between a token of a triple and 
the resources that triple refers to. The reification can be read 
intuitively as saying "'this piece of RDF talks about these things" 
rather than "this piece of RDF has this form"."""

The best health-check here is probably for anyone planning to use 
rdf:predicate, rdf:subject, rdf:object and rdf:Statement to write out 
some test-cases in which the values of these properties are further 
described using owl:sameAs. And then to make sure that substitution of 
equivalents into the graph still fits the intended meaning of the 
extended reification vocabulary.

<rdf:Statement>
  <rdf:subject rdf:resource="http://example.org/doc1.html"/>
  <rdf:predicate rdf:resource="http://purl.org/dc/elements/1.1/source"/>
  <rdf:object 
rdf:resource="urn:lsid:ncbi.nlm.nih.gov.lsid.biopathways.org:genb"/>
  <rdf:object 
rdf:resource="urn:lsid:ncbi.nlm.nih.gov.lsid.biopathways.org:genc"/>
</rdf:Statement>

<rdf:Description 
rdf:about="urn:lsid:ncbi.nlm.nih.gov.lsid.biopathways.org:genb">
  <owl:sameAs 
rdf:resource="urn:lsid:ncbi.nlm.nih.gov.lsid.biopathways.org:genc"/">
</rdf:Description>


Excuse my blundering about in an example from a domain I don't 
understand, the intent was just to give two URIs, such as the LSIDs 
here, that supposedly denote the same thing. So in this example this 
gives us two rdf:object properties of the statement. The roll-your-own 
extensions approach just needs to fit in with this state of affairs, ie. 
in general steer clear of anything designed to talk about the specific 
form of the described RDF fragment...

cheers,

Dan

Received on Monday, 21 May 2007 08:49:23 UTC