W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > May 2006


From: Dan Corwin <dan@lexikos.com>
Date: Thu, 11 May 2006 10:48:56 -0400
Message-ID: <44634ED8.3070108@lexikos.com>
To: Kevin Richards <RichardsK@landcareresearch.co.nz>
CC: swhclsig <public-semweb-lifesci@w3.org>

Kevin Richards wrote:

> I have some questions from a related community.
> The community is a biodiversity inormatics group involved with creating 
> data sharing and interoperability standards (TDWG Taxonomic Databases 
> Working Group - www.tdwg.org <http://www.tdwg.org>)..

Interesting goals.  Those problems do seem familiar.

> What GUIDs are you using within your community, or are intending to use 
> to identify instances of your data (eg PURL, LSID)?

I build domain-specific lexicons for NLP.  Because TDWG
(like me) must work across ontologies, I'd suggest you
consider this "upper ontology" approach.  (Please try
out the URI-creation demo within the default "sandbox"
name space, put there for just that purpose):


Each SAID you create is a GUID *independent* of its name
space.  This makes them handy for bridge ontologies, DB
keys, filenames, etc.  If one SAID gets used in multiple
name spaces, support code outside a reasoner (its importer,
search) can assume all those URIs identify the same concept.

> Will you use the same GUID for all data objects, eg do RDF classes get 
> the same sort of ID as an instance of that class?

Almost, but the creation process differs.  For a class
or property SAID, it takes human thought to tag semantic
facets onto the concept name.  (In theory, this is needed
only for the top-level concepts in domain ontologies.)

But creating a new *instance* URI can be automated if you
get the class SAID - takes only a system clock, flipping
a bit or two, and a new name space.  Collections can use
the same trick, as can a domain's subtypes or prototypes.
So deriving PSAIDs might be made quite inexpensive!

> We are currently looking at several technologies including LSIDs, PURLs 
> and DOIs.  Does anyone have any thoughts on these options?

All above demand a resolution step.  The SAID does not,
giving it several special use cases and benefits.

It does seem compatible with resolvers, however.  If used as
the "object" string in an LSID, for example, resolution of a
whole LSID could be an excellent time to expand the embedded
SAID metadata into normal RDF, exposing it to a reasoner.

Feedback welcome,
Dan Corwin
Received on Friday, 12 May 2006 20:33:15 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:52:26 UTC