- From: Matthias Samwald <samwald@gmx.at>
- Date: Wed, 12 Sep 2007 10:05:11 +0200
- To: Chris Mungall <cjm@fruitfly.org>
- Cc: public-semweb-lifesci <public-semweb-lifesci@w3.org>
> Note that it's impossible to answer the intended query above without > SPARQL-DL - and the most intuitive syntax for this kind of query in > SPARQL-DL may not be triple-based, cosmeticised or not. E.g. "ALL > astrocyte develops_from SOME ?" > > I am heavily biased towards TBox queries - for ABox queries, a > syntactic patina over SPARQL may be very welcome. At the moment I am thinking about ways of expressing subsets of OWL DL in a way that is more easily mapped to the RDF triple model. I simply cannot accept how complicated and sometimes unintuitive the representation of certain OWL constructs in RDF is, although I really tried to. For example, we could have a convention that the triple "Class1 property1 Class2" could be interpreted as as 'some values from' restriction. Of course, not all of OWL could be represented in such a simplified RDF format (e.g., how would we represent 'all values from'?). Yes, I know that we can create OWL APIs and dedicated OWL query languages to make it work. However, I think that technologies should have a certain 'elegance' to find widespread adoption. If we look at some of our HCLS ontologies, we are seeing a XML document that represents a RDF graph in an unnecessarily complicated and error-prone way. And if we look at the RDF graph we see that it represents an OWL ontology in an unnecessarily complicated and error-prone way. A developer that is new to all these technologies might get the impression that the Semantic Web layer cake [1] is just a heap of bad compromises and failed attempts of creating compatibility. [1] http://en.wikipedia.org/wiki/Image:W3c-semantic-web-layers.svg cheers, Matthias Samwald
Received on Wednesday, 12 September 2007 08:05:24 UTC