- From: Bijan Parsia <bparsia@isr.umd.edu>
- Date: Thu, 20 Nov 2003 21:47:48 -0500
- To: public-sw-meaning@w3.org
Hi folks, I got permission from Michael Kifer to forward this to this list. It seems to fit in with some of our discussions. Hmm. at first glance, I thought it might support a "comments count" view, or my own point about the gensym fallacy fallacy, but i see that it might be more against axiomatic approaches. Cheers, Bijan Parsia. Begin forwarded message: > From: kifer@cs.stonybrook.edu (Michael Kifer) > Date: Thu Nov 20, 2003 6:01:24 PM US/Eastern > To: swsl-committee@daml.org (Semantic Web Services Language Committee) > Subject: SWSL declarative semantics > > > > As promised, here is are a few thoughts on the nature of the > declarative > semantics that we need for SWSL. This is a summary of a discussion > that we > had with Karl at ODBASE-03 in Sicily 2wks ago. > > Currently, the General Requirements section states that SWSL should > have > "declarative semantics, in the typical sense used in knowledge > representation where the meaning may be expressed in a > logical framework that establishes overall principles of what > conclusions are sancitoned from a set of premises". > > We felt that in the SW environment, it is inadequate to have semantics > that simply sanctions conclusions. The problem is that a > model-theoretic > semantics by itself doesn't guarantee that all users have shared > understanding of the language constructs and thus use the language > correctly, especially if the logic language at hand is not sufficiently > high level and its semantics is given though a complex set of axioms. > > So, we believe that there is a need for an informal conceptual model > (not > unlike conceptual modeling in databases) that closely corresponds to > the > human perception of objects, classification, processes, etc. (e.g., > UML-like). The language should then have constructs to represent the > concepts of that model directly and the formal model-theoretic > semantics of > these concepts should be natural. By "natural" we mean that a > reasonable > technically competent person should agree that the formalization seems > to > adequately reflect the informal semantics behind the conceptual model. > > In other words, we need to make sure that there is a path from informal > human model of a particular task at hand down to the bowls of the > formalism that underlies the reasoning engine. Human knowledge > engineers > are not going to verify their programs using formal semantics, and in > most > cases they won't even fully understand it. By providing a transition > path > from the informal to the formal we can gain some confidence that the > users' > informal use of the language is reasonably correct. > > This approach applies to Semantic Web in general, not just SWS. > > The above should probably be an objective rather than a requirement. > > > --michael
Received on Thursday, 20 November 2003 21:47:55 UTC