Re: Managing Co-reference (Was: A Semantic Elephant?)

My take on the strength of owl:sameAs issue is that it depends on what 
the application is trying to do.

For some applications, a fairly loose use of owl:sameAs will be really 
helpful, like between:

[[
For example, http://dbtune.org/jamendo/artist/5 and
http://zitgist.com/music/artist/0781a3f3-645c-45d1-a84f-76b4e4decf6d
dentify the same artist. One of them in the Jamendo database, and one
of them in Musicbrainz.
]]

or, what I get the impression to be looser, between

[[
Spain the political entity is the same as Spain the geopolicial region.
]]

or

[[
a city as from Cyc to a wikipedia article of that city
]]

owl:sameAs, like any other predicate is a point of view,
and we can choose to make interpretations of the world in which even 
quite loose notions of identity hold.

If we arrange that owl:sameAs triples of varying strengths are in 
different graphs, then, different applications can load up with the 
strength of sameAs that is appropriate to their needs, their world view, 
their interpretation of the world.

[I am deliberately trying to fudge on whether I mean 'interpretation' in 
a formal or an informal sense]

This sense of their being a multiplicity of world views originated in 
the same collection of named graphs (in the limit, the whole of the 
Semantic Web), is articulated in the named graphs papers in terms of the 
application choosing which graphs to trust, where trust is about fitness 
for purpose, rather than absolute truth.

So, on the example of:
[[
a city as from Cyc to a wikipedia article of that city
]]
a list of all such correspondences, between Cyc and wikipedia, might be 
a great thing to have in a mash-up, and by all means use owl:sameAs.
But keep that list in a separate graph from other data, and I can load 
it for a mash-up, and not for a different application in which such 
loose thinking is not appropriate. Also, the more metadata we provide 
about the metadata we provide, the easier it will be for applications to 
make such choices, so being able to label a graph of somewhat dodgy 
equivalences, as a graph of somewhat dodgy equivalences, in a vocabulary 
that had sufficent deployment, would help.

Jeremy

Received on Thursday, 15 May 2008 17:16:53 UTC