- From: Peter F. Patel-Schneider <pfps@inf.unibz.it>
- Date: Fri, 12 May 2006 17:48:57 -0400 (EDT)
- To: stan.devitt@gwi-ag.com
- Cc: public-rif-wg@w3.org
- Message-Id: <20060512.174857.32194485.pfps@research.bell-labs.com>
In fulfillment of
Peter, apply RIFRAF to SWRL [recorded in
<a href="http://www.w3.org/2006/05/09-rif-minutes.html#action06">
http://www.w3.org/2006/05/09-rif-minutes.html#action06</a>
]
I submit the following response.
Peter F. Patel-Schneider
How SWRL relates to the RAF.
SWRL is an extension to OWL adding unary/binary Horn rules with conjunctive
consequents. SWRL has a number of other features, notably a bunch of built-in
functions, but its defining feature is simply the addition of rules to OWL.
A SWRL rule is thus just a Horn rule (with a conjunctive consequent) where all
the predicates are unary or binary. SWRL forbids consequent-only variables. A
SWRL KB is a collection of SWRL rules and OWL axioms. Predicate names
occuring in SWRL rules have the same denotation as OWL classes and
properties.
Syn: Syntactic Discriminators
1/ Restricted vs. Unrestricted Use of Logic Variables
SWRL does not allow consequent-only variables, but this does not restrict
the expressive power of SWRL in the way one might expect because of the
ability to use OWL classes and properties in the consequents of rules.
This points out a problem with RAF - syntactic variations may or may not
matter, depending on the other aspects of the formalism. RAF appears to
written to characterise rule systems that look like how the authors think
rule systems should look.
2/ Predicate Variables Permitted vs. Not Permitted
SWRL does not allow predicate variables.
3/ Monotonic Lloyd-Topor Extensions Allowed vs. not Allowed
SWRL allows Horn clauses plus conjunctive consequents. It thus allows
some but not all of the Monotonic Lloyd-Topor Extensions. The extension
allowed does not add to the expressive power of SWRL because has a
standard model theory. Several other syntactic extensions would thus
also simply be syntactic sugar. Where, then, does SWRL fit in this
discriminator? Again, there is no indication that parts of formalisms
that don't fit within the "standard" rule view are considered.
4/ Explicit vs. Implicit Rule Scope
SWRL has the notion of a KB or ontology, as does OWL, but ontologies are
not really first-class entities. The placement of SWRL in this feature
space is thus somewhat hard to determine.
5/ Slotted (Keyed, Role-Named) vs. Positional Arguments
SWRL does not have keyed arguments. This, however, is simply a matter of
syntactic sugar. The only real use for keyed arguments is for ease of
human comprehension, something that is of very limited utility in a
representation language of this sort.
6/ Webized vs. non-Webized Names
SWRL uses the same syntactic conventions for syntactic entities as are
used by OWL. It thus fits into the Semantic Web even better than just
using "webized" names.
*/ There are no syntactic discriminators for any of the aspects of SWRL
that tie it to OWL DL.
SeS: Syntactic-entailing-Semantic Discriminators
1/ Homogeneous (Body has Single Expressiveness Class) vs. Hybrid (Body
has Two Expressiveness Classes) Rules
SWRL has precisely the same syntax for antecedents and consequents.
(Well, sort of, as consequents can't have variables that do not occur in
the antecedent.) Where then, does SWRL fit in the above
discriminator? The impact of this discriminator is also dependent
on the meaning of the syntax.
2/ Fact-only (Database Tables) vs. Rule-only vs. Fact-and-Rule Bases
SWRL allows OWL axioms as facts. Where, then, does SWRL fit in this
discriminator?
3/ Function-ful (FO Horn Logic) vs. Function-free (FO Datalog)
SWRL doesn't allow functions. However, SWRL does have equality and
enough machinery to make predicates functional. Where, then, does SWRL
fit in this discriminator?
SWRL incoporates datatypes from OWL and adds some built-in functions.
Does this mean that SWRL has interpreted functions? If the SWRL
built-in functions were removed where would SWRL then fit?
4/ Variable-ful (Non-Ground) vs. Variable-free (Ground) Clauses
SWRL allows variables in rule antecedents and consequents.
5/ Predicate Arity/Arities
SWRL allows only unary and binary predicates. However, SWRL ties them to
OWL classes and properties.
6/ Number of Premises in Rules
SWRL does not restrict the number of atoms in either antecedents or
consequents.
7/ Labeled (Anchored, OIDed) vs. Unlabeled Clauses
SWRL does not label its syntactic entities.
8/ Certain vs. Uncertain Clauses and Atoms
What does this mean?
*/ There are no syntactic-entailing-semantic discriminators for any of
the aspects of SWRL that tie it to OWL.
Sem: Semantic Discriminators
1/ Terminating (all Queries Terminate) vs. non-Terminating (at least
one Query does not Terminate) Rulebases
SWRL is undecidable. However, what does this have to do with semantics
(as opposed to being something about syntax, for example)?
2/ Finite-Model vs. Infinite-Model Rulebases (cf. decidability)
SWRL is silent on whether there can be an infinite number of axioms or
rules.
SWRL does not have a finite model property.
3/ modality/intentionality (beyond FOL)
SWRL is first-order.
*/ There are no semantic discriminators for any of the aspects of SWRL
that tie it to OWL.
Prag: Pragmatic Discriminators
1/ Inference control
SWRL does not have any inference control. SWRL does not even specify
how inference is supposed to work. SWRL only has the notion of
entailment, which is defined model-theoretically.
2/ Complexity
SWRL is a simple language. (Well, at least, the rule part is simple.)
3/ Interoperability
SWRL is silent on this aspect, whatever it is.
*/ There are no pragmatic discriminators that measure how well SWRL fits
into the Semantic Web.
How SWRL relates to the RAF.
SWRL is an extension to OWL adding unary/binary Horn rules with conjunctive
consequents. SWRL has a number of other features, notably a bunch of built-in
functions, but its defining feature is simply the addition of rules to OWL.
A SWRL rule is thus just a Horn rule (with a conjunctive consequent) where all
the predicates are unary or binary. SWRL forbids consequent-only variables. A
SWRL KB is a collection of SWRL rules and OWL axioms. Predicate names
occuring in SWRL rules have the same denotation as OWL classes and
properties.
Syn: Syntactic Discriminators
1/ Restricted vs. Unrestricted Use of Logic Variables
SWRL does not allow consequent-only variables, but this does not restrict
the expressive power of SWRL in the way one might expect because of the
ability to use OWL classes and properties in the consequents of rules.
This points out a problem with RAF - syntactic variations may or may not
matter, depending on the other aspects of the formalism. RAF appears to
written to characterise rule systems that look like how the authors think
rule systems should look.
2/ Predicate Variables Permitted vs. Not Permitted
SWRL does not allow predicate variables.
3/ Monotonic Lloyd-Topor Extensions Allowed vs. not Allowed
SWRL allows Horn clauses plus conjunctive consequents. It thus allows
some but not all of the Monotonic Lloyd-Topor Extensions. The extension
allowed does not add to the expressive power of SWRL because has a
standard model theory. Several other syntactic extensions would thus
also simply be syntactic sugar. Where, then, does SWRL fit in this
discriminator? Again, there is no indication that parts of formalisms
that don't fit within the "standard" rule view are considered.
4/ Explicit vs. Implicit Rule Scope
SWRL has the notion of a KB or ontology, as does OWL, but ontologies are
not really first-class entities. The placement of SWRL in this feature
space is thus somewhat hard to determine.
5/ Slotted (Keyed, Role-Named) vs. Positional Arguments
SWRL does not have keyed arguments. This, however, is simply a matter of
syntactic sugar. The only real use for keyed arguments is for ease of
human comprehension, something that is of very limited utility in a
representation language of this sort.
6/ Webized vs. non-Webized Names
SWRL uses the same syntactic conventions for syntactic entities as are
used by OWL. It thus fits into the Semantic Web even better than just
using "webized" names.
*/ There are no syntactic discriminators for any of the aspects of SWRL
that tie it to OWL DL.
SeS: Syntactic-entailing-Semantic Discriminators
1/ Homogeneous (Body has Single Expressiveness Class) vs. Hybrid (Body
has Two Expressiveness Classes) Rules
SWRL has precisely the same syntax for antecedents and consequents.
(Well, sort of, as consequents can't have variables that do not occur in
the antecedent.) Where then, does SWRL fit in the above
discriminator? The impact of this discriminator is also dependent
on the meaning of the syntax.
2/ Fact-only (Database Tables) vs. Rule-only vs. Fact-and-Rule Bases
SWRL allows OWL axioms as facts. Where, then, does SWRL fit in this
discriminator?
3/ Function-ful (FO Horn Logic) vs. Function-free (FO Datalog)
SWRL doesn't allow functions. However, SWRL does have equality and
enough machinery to make predicates functional. Where, then, does SWRL
fit in this discriminator?
SWRL incoporates datatypes from OWL and adds some built-in functions.
Does this mean that SWRL has interpreted functions? If the SWRL
built-in functions were removed where would SWRL then fit?
4/ Variable-ful (Non-Ground) vs. Variable-free (Ground) Clauses
SWRL allows variables in rule antecedents and consequents.
5/ Predicate Arity/Arities
SWRL allows only unary and binary predicates. However, SWRL ties them to
OWL classes and properties.
6/ Number of Premises in Rules
SWRL does not restrict the number of atoms in either antecedents or
consequents.
7/ Labeled (Anchored, OIDed) vs. Unlabeled Clauses
SWRL does not label its syntactic entities.
8/ Certain vs. Uncertain Clauses and Atoms
What does this mean?
*/ There are no syntactic-entailing-semantic discriminators for any of
the aspects of SWRL that tie it to OWL.
Sem: Semantic Discriminators
1/ Terminating (all Queries Terminate) vs. non-Terminating (at least
one Query does not Terminate) Rulebases
SWRL is undecidable. However, what does this have to do with semantics
(as opposed to being something about syntax, for example)?
2/ Finite-Model vs. Infinite-Model Rulebases (cf. decidability)
SWRL is silent on whether there can be an infinite number of axioms or
rules.
SWRL does not have a finite model property.
3/ modality/intentionality (beyond FOL)
SWRL is first-order.
*/ There are no semantic discriminators for any of the aspects of SWRL
that tie it to OWL.
Prag: Pragmatic Discriminators
1/ Inference control
SWRL does not have any inference control. SWRL does not even specify
how inference is supposed to work. SWRL only has the notion of
entailment, which is defined model-theoretically.
2/ Complexity
SWRL is a simple language. (Well, at least, the rule part is simple.)
3/ Interoperability
SWRL is silent on this aspect, whatever it is.
*/ There are no pragmatic discriminators that measure how well SWRL fits
into the Semantic Web.
Received on Friday, 12 May 2006 21:49:09 UTC