W3C home > Mailing lists > Public > public-owl-dev@w3.org > July to September 2008

Alternative interpretation of wrongly instantiated object properties

From: Denny Vrandečić <dvr@aifb.uni-karlsruhe.de>
Date: Tue, 30 Sep 2008 15:53:43 +0200
Message-ID: <48E22F67.6070706@aifb.uni-karlsruhe.de>
To: Owl Dev <public-owl-dev@w3.org>

just a thought, and I guess it has a number of problems:

Assume an object property O and a property assertion

a O lt

with lt being a literal and a an individual.

Now this would yield an inconsistency since O is an object property, and
 lt a literal. But if instead in OWL2

a O lt

would semantically mean

a O x . x rdfs:label lt

with x being a new individual, this would not be an inconsistency
anymore and it would capture the meaning of most cases this occurs.

To give an example, assume some Semantic Web beginner claiming in his
FOAF-file the following:

my:me foaf:knows "Dan Brickley".

it is obviously a mistake, since foaf:knows is an object property
connecting two persons, and not a person with a literal (the same
happens with DC-Vocabularies and other widely used vocabularies).

The actual reading would be

my:me foaf:knows _someone .
_someone rdfs:label "Dan Brickley" .

that our beginner knows someone with the name Dan Brickley (which is
indeed a literal).

There are a number of problematic technicalities this leads to:
1. the triple my:me foaf:knows "Dan Brickley" does not disappear, and
due to the downward-compatibility to RDF it still exists, which means
that asking for my:me foaf:knows ?x gives different answers depending if
you had an OWL2 aware SPARQL-engine instead of one that only does RDF.
And they do not grow monotonically by adding OWL2 -- I am afraid this is
a problem. (but then again, the relationship of OWL2 and SPARQL is not
very well defined yet, but I assume that such a form of monotonicity
would be highly desirable)
2. this introduces new individuals, and I am not sure if this has some
theoretical implications.

The positive aspect is that it would increase the percentage of valid
triples. The above stated error appears quite frequently.

Then again, it could also be just best practice to "repair" an OWL- or
RDF-document in such a way before actual processing, just as
HTML-documents are usually repaired before a DOM is being created or
before the document is being rendered. But then again, repair best
practices seem to become standardized for HTML5, and thus it may be an
idea to pursue this for OWL2 as well (probably as a separate document).

Received on Tuesday, 30 September 2008 13:54:18 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:58:17 UTC