Metathesaurus issues. was: Re: does RDF require understanding all 82 URI schemes?

David Megginson wrote:

> Martin Duerst writes:
>
>  > The answer to this was obvious to me two seconds after reading
>  > through your complaints. And then in the next mail showed that
>  > it already exists:
>  >
>  >  > <#x>
>  >  >   daml:equivalentTo
>  >  >   http://www.megginson.com/battles.rdf#jutland
>  >
>
> I already answered this one privately -- this solution won't scale,
> because it requires you to retrieve a resource to find the aliased
> identifier.  If you're dealing with more than a few objects, you'll
> end up with an exponential explosion of network access.  That's not to
> say that an equivalent-to property isn't a good idea; only that it's
> not a practical alternative to sharing the same identifier.

Alternatively, and preferably in many cases, construct a thesaurus. Perhaps
this is an application for topic maps. Perhaps an application might extract
daml:equivalentTo properties into an external document.

>
>  > Do they need that? Maybe not. In everyday language, we never
>  > use social security numbers. The identity of the objects we speak
>  > about is part of the web of language, not outside of it.

In healthcare the SSN is the closest (yet imperfect) unique identifier.
Several problems:

1) SSN's are reused (so same problem as mailboxes).
2) People sometimes 'share' a friend or relative's SSN... this is the real
killer for any unique identification system other than a biometric
(fingerprint or retinal scan ... these can also change though!)

But this problem is a huge one for the healthcare field. Things like "master
patient index" are good examples of real world thesauri which map so-called
unique identifiers (e.g. patient id) from one part of an institution to
another. Having taken part in the merging of several healthcare
organizations I can assure you that its all alot of fun (and good healthy
human readable prefixes to UIDs such as are found with URLs are a really
huge help in these cases).

Jonathan Borden
The Open Healthcare Group
http://www.openhealth.org

Received on Tuesday, 13 February 2001 10:17:18 UTC