From: Mike Dean (mdean@bbn.com)
Date: 12/01/03
Here's the input I promised during last week's telecon.
I think we should continue to take small steps in evolving
SWRL. There are 2 related features that I think would give
users much-needed capabilities and make a nice follow-on
language increment: datatype comparisons and arithmetic
conversions.
Many rule examples (including the financial examples given
by Said at the DAML PI Meeting [1]) involving numeric
comparisons, e.g.
// free shipping on orders of $50 or more
if (order.cost < 50)
order.shipping = 5
else
order.shipping = 0
OWL inherits DAML+OIL's practice of using XML Schema
Datatypes to define data ranges, e.g. defining a SmallOrder
to be an Order with cost user:lessThan50. Even aside from
uncertainties in URI naming of such user defined types, this
is awkward and has not been widely implemented. There's a
good opportunity for SWRL to solve this problem.
My preferred solution is to incorporate support for the
ordered XML Schema datatypes used by OWL directly in SWRL,
e.g. by introducing a ComparisonAtom that takes 2 d-objects
and is either "subclassed" (e.g. LessThanAtom) or takes an
operator named by URI. While there's a temptation to treat
such operators as one case of procedural attachments, I
think that adds complexity and would prolong adoption.
Primitive datatype comparisons are handled directly by most
rule systems.
My second proposed addition is support for arithmetic
datatype conversions sufficient to write translation rules
(e.g. converting ont1:fahrenheitTemperature to
ont2:celsiusTemperature, or ont1:priceInDollars to
ont2:priceInEuros based on the current exchange rate). The
basic arithmetic operations are the most critical - support
for a larger set of side-effect free numeric and string
operators (e.g. cosine, string concatenation, and case
conversion) such as the XQuery builtins could be desirable
if it doesn't add significant complexity or delay. Again,
such operators could be viewed as part of a more general
procedural attachment mechanism, but I don't think we're
ready for that.
Mike
[1] http://www.daml.org/2003/10/pi-meeting/rulemlight-talk_Boley.pdf
This archive was generated by hypermail 2.1.4 : 12/01/03 EST