[redland-dev] RDFS/OWL and reasoning in Redland
Richard Newman
r.newman at reading.ac.uk
Wed Sep 27 19:04:38 UTC 2006
> All right, that is a different, butvalid point of view fromI =
> wanted. Allowing anything to be stated with validation after the =
> fact. Nevertheless, there does not seem to be a facility to allow =
> for checking the validity of statements according to an ontology. =
> What would you suggest?
That's not how the semantic web works.
You can say anything. Everything you say might cause additional =
information to be inferred. That additional information, or the =
statement itself, might cause your knowledge base to become =
inconsistent. (The Semantic Web as a whole is almost certainly =
inconsistent, if only because of Danny's cats' FOAF profiles!)
The store itself can't know which statement =97 the one already in the =
store, the new one, or one of their logical consequences when =
combined =97 is wrong; it could even be a flaw in one of your ontology =
definitions. It knows only that the store was consistent, and now is =
not.
If you wish, you could use some kind of wrapper:
(defun add-triple (triple)
(let ((old-store (copy-store)))
(add-triple-to-store triple)
(when (inconsistent? store)
(error "Store would be made inconsistent by adding this triple.")
(replace-store old-store))))
(obviously implemented more efficiently, but you get the idea) but =
that doesn't give you much to work with. Here's an example of why:
Store A:
Richard is a Car
People can't be Cars
Add 'Richard is a Person'. That statement, which is not intuitively =
invalid, causes an error.
What would I suggest? Use a tool such as Schemarama, or even basic =
queries, to check that data is there (that Person X has a birthday =
and a name). Use a tool such as Racer to ensure that your data is =
consistent (that Person X isn't simultaneously a Car and a Person). =
And stop worrying about validation of your RDF.
-R
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.usefulinc.com/pipermail/redland-dev/attachments/20060927/=
6f71ea4c/attachment.htm
More information about the redland-dev
mailing list