|We warmly support the introduction of language profiles, and we think
they are on the whole well chosen. However, we find it unacceptable that the only way to determine the language fragment of an ontology is to ... parse the entire ontology. This is clearly impractical. An obvious usage scenario is that (1) I'm looking for ontologies in a particular application domain (2) I can only tolerate ontologies in/up-to a particular profile
(eg. because of limitations of my processing power)
Doing (1) can be done by something simple like treating all candidates as text documents and look for relevant keywords, but for (2) .... I must parse all candidate ontologies to see if they satisfy point (2). (not nice, when the the OpenCyc ontology is one of my candidates...) And what's worse, everybody has to do this again, because there is no standardised way of expressing my findings anywhere. Why not choose the obvious option to introduce a designated AnnotationAssertion, somewhere in the header of the ontology document, to state the (minimal) language profile of the content, by way of a designated URI for each language profile. Of course there is no guarantee of correctness for those annotations, but neither is there for any other statement that an ontology makes. We find the entire discussion on issue 111 <http://www.w3.org/2007/OWL/tracker/issues/111> and its outcome unconvincing and strongly urge to reconsider.
Frank van Harmelen, and many members of the Semantic Web Group at the Vrije Universiteit Amsterdam
PeterPatel-Schneider 04:49, 14 February 2009 (UTC)
[Response for LC Comment 31]
Dear Frank (and many others),
Thank you for your message
on the OWL 2 Web Ontology Language last call drafts, which the WG has taken to concern adding a syntactic marker for the expressive power of used in an ontology, including which profile the ontology belongs within.
As you point out, this kind of a marker was discussed in relation to WG issue 111 (see http://www.w3.org/2007/OWL/tracker/issues/111). We are sorry that you find the discussion related to Issue 111 unconvincing - after this discussion the WG decided that markers were in general not a good idea. In this message I have included some points related to syntactic markers.
One problem with having an expressive power marker in the header of an OWL ontology is that there are syntaxes for OWL ontologies, in particular the preferred syntax, RDF/XML, where the header information is mixed in with the rest of the ontology. Recovering the ontology header in such syntaxes can thus require parsing the entire ontology. Note that simple text processing is not adequate as the marker syntax could occur in places where it should not be interpreted as the marker, e.g., in comments or not attached to the ontology node.
Another problem is how to treat cases where the expressive power of the ontology does not match the marker. This is very different from what to do if an ontology does adequately reflect reality. Should tools be required to reject ontologies where the expressive power of the ontology is greater than what the marker states? Should tools be required to reject ontologies where the expressive power of the ontology is less than what the marker states? Should nothing be said? What guarantees should a search for expressive markers provide? The answers to these questions are not obvious.
The current situation leaves open the possibility that users will get together and, through practice, provide the answers to these thorny questions. Ontology annotations can be used to support this effort. If a body of practice arises, then it may be appropriate for a subsequent working group to consider adding this practice to OWL.
In light of the discussion of Issue 111 and the reasons stated above the OWL WG does not intend to make any changes in response to your comment.
Please acknowledge receipt of this email to <mailto:email@example.com> (replying to this email should suffice). In your acknowledgment please let us know whether or not you are satisfied with the working group's response to your comment.
Regards, Peter F. Patel-Schneider on behalf of the W3C OWL Working Group