On Sat, 17 Jul 2010 12:02:19 +0100 Richard Light <[email protected]> wrote:
> If [a museum] were to follow the dbpedia model, and publish a set of > [unrelated] triples with the object identifier as subject, embedded > in their web page for the object, there is nothing to stop someone > else putting out a page containing lies about that object, also > expressed as simple triples with the object URL as subject. There's also nothing to stop people publishing a page containing the truth about an object if they notice that the museum has told lies about it. > By the time Google has indexed both those pages "semantically" (see > yesterday's acquisition of FreeBase) and merged the results in its > uber-index, you won't know the difference. That assumes a fairly naive approach to graph management. Most large-scale RDF stores keep statements as quads (or even quints), giving you a slot to record the provenance of each piece of information. The consumer can then apply as sophisticated a policy as they like to decide which sources they trust. -- Toby A Inkster <mailto:[email protected]> <http://tobyinkster.co.uk>
