From: Sandro Hawke (sandro@w3.org)
Date: 06/04/03
I think I heard two complaints about Benjamin's proposed style of
rules-language semantics:
1. It's different from RDF and OWL, and that difference itself
makes things more complicated. Such a difference needs to
be well-motivated.
2. It's wrong, because it means you can't express knowledge about
real things, like specific colors.
I would like to see a test case for #2. Pat sketched one using
colors, but I couldn't quite see the details. Let me try another one
using Said's example of financial regulations being published as
Rules.
As I recall from the bubble days, the American SEC has a rule saying
that company founders cannot sell any stock within 6 months after the
company goes public. Perhaps they could publish a rule which
various consumers could incorporate into their stock-trade planning
software:
sec:forbiddenTrade(?x) <-
sec:seller(?x, ?person) /\
sec:company(?x, ?company) /\
sec:date(?x, ?saledate) /\
sec:founder(?person, ?company) /\
sec:ipoDate(?company, ?ipodate) /\
% ?saledate < ?ipodate+6 months; requires some date handling
Let's look at that "sec:founder" predicate. In RDF's semantics I
would say it relates a person and a company. In HornLP semantics, I
would say it relates a term which presents a person and a term which
represents a company. Is that right? Now the question is: how do we
get the difference to show up? Where does it matter?
Benjamin's notes on the subject say:
> Benj: that's orthogonal to Herbrand-ness unless you can represent within
> the KR some info relating the denoter to the denotation
but I can't see how to get to that situation where it stops being
orthogonal.
-- sandro
This archive was generated by hypermail 2.1.4 : 06/04/03 EST