[rdfweb-dev] Smushing the Semantic Web / Invalid data in IFP's

James Carlyle james.carlyle at takepart.com
Mon Mar 22 11:37:23 UTC 2004


Perry

Scutters that smush should adopt the most careful strategy, to avoid
false assertions being presented later.  A conservative smushing
strategy is to ignore smushing of resources that have contradictions in
properties that are specified as unique.

For example, both foaf:jabberID and foaf:mbox_sha1sum have OWL Type: An
InverseFunctionalProperty (uniquely identifying property).  

So we would expect a pattern of 

Person
	jabberID 		"jc at jabber.dk"
	mbox_sha1sum	"abc123"
Person
	jabberID 		"iand at jabber.dk"
	mbox_sha1sum	"def456"

If we got contradictions, like two Persons who had different uniquely
identifying mailbox hashes, but the same uniquely identifying jabberIDs,
we could abandon smushing on those resources.

Person
	jabberID 		"wtf is jabber" - clearly not a
"uniquely identifying property"
	mbox_sha1sum	"abc123" - clearly a different Person
Person
	jabberID 		"wtf is jabber" - clearly not a
"uniquely identifying property"
	mbox_sha1sum	"def456" - clearly a different Person

Best regards,

James Carlyle
www.semanticplanet.com




More information about the foaf-dev mailing list