From: Drew McDermott ([email protected])
Date: 02/11/04
Sandro has mentioned a couple of issues that come down to whether DRS
is based on the assumption that a document encoding a formula in RDF
using the DRS hack must be taken as a whole, with no guarantees that
subtracting some arbitrary bunch of triples will leave a document
whose meaning is entailed by that of the original document.
More clearly: Let D be a set of triples obtained by using the DRS/SWRL
translation mechanism to a logical formula F. Let D' be an arbitrary
subset of D, which happens to be the DRS/SWRL encoding of formula F'.
Sandro points out that F may not entail F', which is contrary to the
spirit of RDF.
I have to plead guilty. Making the triple be the fundamental unit of
meaning in RDF was a bad idea, and no reasonable solution to the KR
problem on the web can retain it. The fundamental unit of "speech
acts" in all other systems is defined by the syntax of a language.
In most logics, the "speech acts" include assertion (always), plus
querying, sequent posting, etc. The syntax of the language determines
what can be legally asserted, queried, etc. To use DRS, you start
with a language that defines these categories of well-formed formulas,
you translate into RDF, and insist that if an agent is asserting a
document representing a well-formed formula, then nothing can be said
about whether it is asserting other well-formed formulae that might be
extracted from the document.
-- Drew
This archive was generated by hypermail 2.1.4 : 02/11/04 EST