RE: BioRDF [Telcon]: slides for the UMLS presentation

Hi Ben,

Thanks for your thoughtful reply.

> 1) Accessible URIs for UMLS terms would make it possible (easier..)  
> for any one to annotate their data through reference to this 
> expertly  
> curated common vocabulary.  This makes it much more likely that  
> people would actually do so - thus enabling serendipitous/automatic  
> data integration down the road.

LSIDs provide this without the need of translating to an imperfect OWL.
Even if there is no true LSID resolver service for UMLS terms, one can
decompose the LSID and use the varying existing URI to access directly
the information on UMLS terms.

> 2) Even without using the semantics of the relationships, the simple  
> fact that certain concepts are somehow associated with each other is  
> excellent fodder for computation and exploration.

Yes, I agree.  This might be the most compelling reason.  But how much
can ontology reasoning tools use the specialized relationships in the
OWL representation to draw inferences?  If it is purely statistical on
the standard OWL relationships, many additional linkages will be missed
and even those that are found, the significance will only be apparent by
going back to the source.

> 3) By providing RDF, it would be possible to automatically merge RDF  
> generated in other projects with the UMLS.

True, and I agree the statistical possibilities of integration expands. 

My main point is what role has these specialized relationships for UMLS
OWL added in the above scenarios?  Why not stick with straight OWL
representation and do any further analysis from the UMLS source?  My
complaint is that the expansion of the OWL representations to have these
special case relationships obscures the benefits of OWL. 

cheers,
Michael

> -----Original Message-----
> From: Benjamin Good [mailto:goodb@interchange.ubc.ca] 
> Sent: Monday, June 05, 2006 10:45 PM
> To: Miller, Michael D (Rosetta)
> Cc: olivier@nlm.nih.gov; matt@biomedcentral.com; public-semweb-lifesci
> Subject: Re: BioRDF [Telcon]: slides for the UMLS presentation
> 
> 
> 
> On Jun 5, 2006, at 1:04 PM, Miller, Michael D (Rosetta) wrote:
> 
> > Hi Ben and Matt,
> >
> >> An RDF version of the
> >> UMLS knowledge sources would be seem to be very useful - at
> >> least for
> >> bioinformatics research purposes ...
> >
> > What I was trying to discover was, from a purely RDF/ontology
> > standpoint, what is gained without any other knowledge of UMLS?
> 
> 1) Accessible URIs for UMLS terms would make it possible (easier..)  
> for any one to annotate their data through reference to this 
> expertly  
> curated common vocabulary.  This makes it much more likely that  
> people would actually do so - thus enabling serendipitous/automatic  
> data integration down the road.
> 2) Even without using the semantics of the relationships, the simple  
> fact that certain concepts are somehow associated with each other is  
> excellent fodder for computation and exploration.
> 3) By providing RDF, it would be possible to automatically merge RDF  
> generated in other projects with the UMLS.
> 
> Did I miss any ?
> 
> >
> > This question comes more from my ignorance, probably, but when  
> > knowledge
> > gets buried in specialized tags the tendency is that 
> without going to
> > the source to understand the specialization, what's left is 
> not all  
> > that
> > useful.
> 
> There is no question that you would get maximum value out of a  
> thorough understanding of the meaning of all of the relationships in  
> the RDF and that the farther we can push this understanding into an  
> ontology (or other reasonable representation) the better; 
> but, moving  
> things like the UMLS into RDF and onto the Web IMHO increases their  
> value to the community tremendously and is a necessary step towards  
> actually realizing the semantic web for the life sciences (not A  
> semantic web for the life sciences).
> 
> -Ben
> 
> 
> 
> 
>

Received on Tuesday, 6 June 2006 15:06:46 UTC