Short motivation for

Proposed OWL Knowledge Base Language

Frank van Harmelen
Peter F. Patel-Schneider
Ian Horrocks

19 March 2002

Table of contents

Preliminary remarks

Our proposal for the OWL Knowledge Base Language is given in an abstract syntax. This syntax identifies the language's constructions: the basic terms of the language, and how they can be combined. The precise surface syntax(es) remains to be determined.

Our grammar may be ambigous in a few places. We consider this unimportant at this stage. We aimed for a readable grammar that could be the basis of discussion on the WG. Ambiguities etc. will be ironed out when defining the surface syntax(es).

We believe that this proposal is in line with our first sketch of OWL's language features, which was agreed upon during the 7 March teleconf. We also believe it to be consistent with our charter's requirement to take DAML+OIL as the basis for OWL.

Core of the proposal

Our proposal for the OWL Knowledge Base Language comes in two parts:
  1. The first ("light") part is loosely based on the frame idiom found in the frame-style systems that have been used in AI for decades. This idiom has been extended with commonly found ontology modelling idioms and a number of features that are important in the Web context.
    This "light" version will provide a lower entry threshold to the language, while still providing much of the required expressiveness.
  2. The second ("full") part is very close DAML+OIL. Since this part is already well understood, our comments concentrate on the "light" segment of our proposal.

The frame portion of the definitions captures the widely understood frame idiom.

Additional remarks

  1. Notice that the "light" idiom only uses named classes for superclasses, domains, ranges, etc. The use of arbitrary class-expressions is still possible in full OWL.
  2. Notice that the "light" part includes all the RDF Schema idioms.
  3. Notice that the "light" idiom limits cardinality constraints to the values 1 and "greater than 1" (as mathematicians say: "0,1,infinity" :-)
  4. Notice that the (in)equality statements for individuals, classes and properties are all n-ary, for compact expressiveness.
  5. Notice that our proposal only deals with the "knowledge-base language" part of OWL. Requirements like tagging, importing ontologies, etc are not dealt with.