Notes
Slide Show
Outline
1
A Web Rules WG
Charter Focus
Strawman Proposal

Version 1.1, April 30, 2005

This Version Prepared by:
Benjamin Grosof, Harold Boley, Michael Kifer, and Said Tabet
of The RuleML Initiative (http://www.ruleml.org).
Incorporating Comments by Ed Barkmeyer.

***Further revisions to be incorporated from community discussion.***

Modified from earlier version in RuleML Position Paper [96] of the
W3C Workshop on Rule Languages for Interoperability, 27-28 April 2005.
Responsive to the discussion from that Workshop, and from the WSMO-RuleML-SWSI Face-to-Face Meeting of 26 April 2005.

                                                                          “WG” above = W3C Working Group
2
The Web Rule Language in its Context
3
Semantic Interoperability Principles - high level
  •  Conclusions sanctioned do not depend on how executed, e.g., forward chaining has same semantics as backward chaining
  •  “Reaction” rules, that perform side-effectful actions, have a semantics which cleanly extends the basic case of rules that do not.
4
Focus Overall of WG
  •  Kernel based on logical KR
    •  Semantics, syntax, layering:  for that kernel
    •  Rudimentary rule management:  e.g., queries, answers, premises, conclusions, updates to premises, ruleset definition, importation of rules, simple versioning, simple provenance
  • Use Cases from Business Processes, Services
    •  Policies, in particular
    •  Support Semantic Web Services requirements, in particular
  •  Integrate Rules and Ontologies
    •  Interoperate with OWL, in particular
    •  Represent Ontologies as Rules, in particular
5
Rule Communities Served
  •  Semantic Web
    •  general, using XML and/or RDF encoding
    •  RDF- and OWL-centric, in particular
    •  Logic Program based, in particular
  •  Business Rules
    •  general, based on existing rule-based
    •  Production Rules, in particular
6
Kernel KR Focus
  • Declarative Logic Programs expressiveness including
  •   1. Datalog Horn LP     (N-ary predicates supported)
  •   2. + scoped default negation applied to atoms
  •           a. simple extensional
  •             b. more general (allowing inferential chaining to establish the
  •                                                           atom in question -- subset of, or full,
  •                                                           Well Founded semantics)
  •   3. + procedural attachments (external calls)
  •             a. actions (side-effectful – external)
  •             b. tests (side-effect-free queries)
  •   4. + logical functions, incl. for object creation, skolemization
  •                    a. limited initially (to ensure finite/tractable forward inferencing)
  •                    b. more general (e.g., for backward chaining, “sugar” features)
7
Kinds of Rules & Rule Systems
Translatable/Reducible to Kernel
  • Most other wish-list features
  •   can be expressively reduced to this core KR abstraction, for which Situated Ordinary Logic Programs can provide the semantics theory


  • OWL:  large subset, Þ
  • OWL ontology integration via overlap of LP with Description Logic
  •                  (e.g., use Description Logic Programs V2,
  •                                with integrity constraints, skolemization, equality, passing of derived facts)
  • SWRL:  large subset
  • Production Rules cf. PRRuleML:  large subset (Production
  •                                                                                             Logic Programs)
  • Decision trees
  • Decision tables
  • “Sequential” rules cf. PRR:  [**probable, need to understand better]
  • Prolog:  the pure subset (which is large)
  • SQL relational databases:  large subset (incl. all core)
  • Event-Condition-Action rules:  large subset


8
Those are translatable/reducible because the following are …
9
Additional “Sugar” Features that are
Translatable/Reducible to Kernel
  • Most other wish-list features
  •   can be expressively reduced* to this core KR abstraction, for which Situated Ordinary Logic Programs can provide the semantics theory          (* with tractability, known techniques).
  • E.g., much or all of the expressiveness in the following.


  • RDF facts
  • Frame syntax
  • Slotted syntax
  • Lists
  • (N-ary predicates if restrict core to 2-ary)
  • RDFS-DL simple ontologies
  • Datatyping:  basic
10
Sugar Features II
  • “Else” part of if-then-else
  • Courteous prioritized defaults,
  •                      incl. declarative priorities, limited strong/classical negation,
  •                            prioritized conflict handling, paraconsistency robustness
  • Default inheritance cf. Object Oriented programming, “frame” languages
  • “Hilog” – quasi higher order syntactic sugar
  • Lloyd-Topor
  • Integrity constraints that report violations
  • Anonymous existentials, blank-nodes, limited skolemization
  • “Crud” – create update delete, cf. Production Rules   (restricted)
  • “Assert”, and basic “retract”, cf. Production Rules  (restricted)
11
Sugar Features III
  • Reification, basic
  • User equality, basic aspects
  • Equations, basic
  • Built-ins (side-effect-free functions/operators, read/write)
  • Access to surrounding object-oriented data environment, cf. OO Production Rules
  • Ontological context translation & mediation
  • Contextual selection conditions for whole rulesets
  • “Rules flow”:  some  (e.g., sequencing of rule groups)


  • … probably some more things we forgot to list here …
12
The Web Rule Language in its Context
13
Layering Relationships wrt existing Semantic Web Standards
  •          subsumes (expressively)
  •          layers-on (makes use of)
  •          overlaps-with (expressively)
14
Sugar Features vs. Kernel
  • “Sugar-enhanced” Languages can be translated into the kernel.
    • I.e., Sugar Features can be implemented/supported via translators
    • Including as “best practice”, etc.

  • Could consider doing some of them as part of WG proper
    • E.g., basic set of datatypes
  • … But it’s not as crucial
15
Deliverables Desired
  • Abstract syntax
  • Semantics
  • Layering definitions:  e.g., Datalog Horn layer
  • Concrete syntax:
    • Markup syntax in XML
    • RDF (e.g., RDF/XML)
    • Human-readable presentation (non-XML) syntax
  • UML/MOF metamodel
  • Some light ontology about rudimentary rule management, incorporated into the above
    • E.g., to enable representing provenance, or expressive restrictions met, about a particular rulebase
16
Supported Tasks & Kinds of Knowledge
  • Policies:  authorization, contracting, security, privacy, monitoring, advertising, regulations, governance, …
  • Validation:  integrity, notification, …
  • Business Processes, Workflows, Protocols, …
    • Process modeling:  Abstract State Machines, Pi-Calculus, …
  • Semantic Web Services
  • Ontologies
  • Mediation:  map between ontologies/contexts
  • …