RE: HL7 and patient records in RDF/OWL?

1) RDF is a semi-structured data model itself.
	A: I agree and in fact part of the BIORDF group's effort will be to
evaluate 'semantics requirements' of the resulting RDF (see Stage 2, task 2
in the BIORDF proposal)
	
  

2) A relational database is extremely structured.
	A: Yes, but, as I see it, semantics is not formalized nor,
frequently, explicit in relational models (E-R relationships are not
necessarily first class objects, see the example below).  
	

3) The difficulty of transforming a semi-structured data-base into an
RDF/OWL knowledge-base depends on whether or not the target ontology is
pre-specified.
	A: Right. I think the question is: when do you need an ontology?
(semantic requirements). Probably, as you point out, you need it in the
first place, but there predictably some simple data that can be mapped into
each other without an ontology.
	


>'For example, transforming a database into OWL is trivial.  Every
>relation is a class, every tuple is an instance of that class.  Every
>attribute of a given tuple is a triple connecting the tuple to its
>value for that attribute.'

  
I'm not sure is always that easy. Relations in data-bases can be, for
instance, represented with table values (imagine a table containing a field
'relation' accepting values such as 'is_a', 'part_of', 'causes', increases'
etc). 



>'Transforming arbitrary XML into OWL is
>described by Sergey Melnik [1].  (Unfortunately, Sergey seems to have
>lost his Stanford website.)'


Thanks for the reference!

>On the other hand, if you have a specific target ontology in mind, then
>the problem is at least as hard as any semi-structured mapping problem.
>I say at least as hard because you must decide when to jump formalisms,
>then solve the integration problem in that formalism.

Right


Davide

Received on Friday, 10 February 2006 17:48:12 UTC