- From: Peter Mika <pmika@cs.vu.nl>
- Date: Wed, 2 Jun 2004 14:10:32 +0200
- To: <public-swbp-wg@w3.org>
Dear All, (Apologies in advance for the potential misuse of the mailing for providing public comment. I'm not a member of the WG, but I failed to locate a public comment mailing list.) I have read the Classes as Values note (version 3, by Natasha Noy) with great interest as the representation issue of 'annotation with subjects' have been a topic of discussion in several projects I was involved in. In every case, the advantages and disadvantages of various approaches have been discussed and then a decision was made for one or the other. For example, in the EU FP5 project On-To-Knowledge we settled for Approach 2, except that we used our own property called isAbout (instead of dc:subject) and added the rule that (Doc, isAbout, A) and (A, rdfs:subClassOf, B) -> (Doc, isAbout, B) which provided the missing piece of inference (not expressible in OWL DL, as rightly noted in the document). Despite that the note is a clear and important summary of these discussions and is, for example, already read with interest by members of the EU FP6 SEKT project, I personally have the feeling that merely listing the various approaches that one can take does not go far enough when it comes to the Semantic Web. Within a project or in a closed community one may choose any representation that he sees fit as long as it's agreed, but the case of the Semantic Web is somewhat different. I suggest that this group should consider what happens if we allow people to choose from these options at will and annotate their content correspondingly. If I let my crawler out on the Semantic Web, how is it going to figure out on the first place which approach is being used? (Especially with the multiple interpretations given to dc:subject...) Is it still possible then to integrate metadata created with the different approaches? What are the translation rules for that? In summary, I have the feeling that the level of commitment this document represents is still too low to build an interoperable Semantic Web of content and annotations. My suggestion to the group is to either (1) limit the choices further or (2) to describe the ways to identify the various approaches and how to make them interoperable (translate them to one another) with minimal loss of semantics. Best, Peter ----------------------------- Peter Mika M.Sc. Free University Amsterdam (VUA) Faculty of Sciences Business Informatics Section De Boelelaan 1081a 1081HV, Amsterdam Tel.: +31 20 4447452 http://www.cs.vu.nl/~pmika -----------------------------
Received on Wednesday, 2 June 2004 08:11:04 UTC