- From: Richard Fikes <fikes@KSL.Stanford.EDU>
- Date: Thu, 15 Aug 2002 22:25:14 -0700
- To: www-rdf-logic@w3.org
- CC: Deborah McGuinness <dlm@KSL.Stanford.EDU>, "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>, macgregor@ISI.EDU, Ian Horrocks <horrocks@cs.man.ac.uk>
Here are some additional comments on this topic -- It is very useful and not unheard of for representation systems to include classes and predicates as first class objects. Prominent examples of which I am familiar include the versions of the Knowledge Interchange Format (KIF) that included set theory (e.g., version 3.0, http://www-ksl.stanford.edu/knowledge-sharing/kif/) and the OKBC knowledge model (http://www.ai.sri.com/~okbc/). In those versions of KIF, a predicate is a set of lists, all finite lists of objects in the domain of discourse are in the domain of discourse, and all sets of objects in the domain of discourse are in the domain of discourse. In the OKBC knowledge model, all classes (which are unary predicates), all slots (which are binary predicates), and all facets (which are ternary predicates) are in the domain of discourse. Including predicates and classes in the domain of discourse provides very important expressive power for many domains and applications. For example, it enables one to define classes of classes (e.g., the class of "year-months" whose instances are the classes "January", "February", etc., each of whose instances is a month of a particular year), to assert property values for classes (e.g., their cardinality), and to assert properties of predicates (such as being transitive or symmetric). I would think that many uses of the Semantic Web will require the expressive power provided by treating classes and predicates as first class objects. Richard
Received on Friday, 16 August 2002 01:25:52 UTC