The RuleML Initiative, July/August 2001 International Semantic Web Working Symposium SWWS-01 30 July - 1 August 2001, Stanford, California, USA Harold

The RuleML Initiative, July/August 2001

International Semantic Web Working SymposiumSWWS-01

30 July - 1 August 2001, Stanford, California, USA

Harold Boley, Said Tabet and Gerd Wagner

Design Rationale of RuleML: A Markup Language for Semantic Web Rules

The RuleML Initiative, July/August 2001 2

Introduction and Background The RuleML Initiative as a Web Ontology Effort Modular Syntax & Semantics of RuleML RuleML DTDs: release 0.8 Negation Handling in RuleML Priorities/Evidences in RuleML RuleML Implementations via XSLT & Rule Engines Example: A Semantic Web Scenario in Insurance Conclusions

Agenda


Introduction and Background

Semantic Web rules less studied than corresponding ontology (taxonomy) markup

Rule Markup Initiative fills this gap by exploring rule systems (e.g.,extended Horn logics) suitable for the Web:

Syntax (XML and RDF), semantics, tractability/efficiency, transformation, and compilation

• Derivation rules: may be evaluated bottom-up as in deductive databases, top-down as in logic programming, or by tabled resolution as in XSB• Reaction rules: also called "ECA" -- “Event-Condition-Action" -- or “Trigger" rules• Possible combinations

RuleML


The RuleML Initiative as a Web Ontology Effort

RuleML Initiative: started in August 2000 during the Pacific Rim International Conference on Artificial Intelligence (PRICAI 2000)

February/March 2001: First RuleML BOF at W3C Plenary and WG Meeting event

Ontology ‘equation’: Ontology = Taxonomies + Axioms(with Axioms having the form of Rules) gives big picture

System of sublanguages: from RDF-like triples to a labeled Horn logic with equations plus URI individuals and relations


Modular Syntax & Semantics of RuleML

Reaction Rules

1. 2.

Integrity Constraints Derivation Rules

3.

Facts

RuleML hierarchy top-level:

1. Integrity constraints: reaction rules that signal inconsistency when certain conditions are fulfilled

2. Derivation rules: reaction rules whose action happens to only 'assert' a conclusion when certain conditions (premises) are fulfilled

3. Facts: derivation rules with empty conjunction of premises


Modular Syntax & Semantics of RuleML (Cont’d)

urc-bin-data-ground-facturc-bin-data-ground-log

urc-bin-datalog

bin-datalogurc-datalog

ur-datalog

ur-hornlog

ur-equalog

hornlog

equalog

datalog

ur

Rooted DAG will be extended withbranches for further sublanguages

URL/URI-like‘ur’-objects

derivation rules

RDF- like triples

ur-datalog = join(ur,datalog

)

Sublanguagehierarchy:

RDF Rules


Modular Syntax & Semantics of RuleML (Cont’d)

Separation of concerns:Sublanguage hierarchy:

12 sublanguages together constitute modularized basic RuleML definition. Except 'UR' (URL/URI) group correspond to well-known rule systems, each with semantic (model- and proof-theoretic) characterizationConcrete markup:

In recent months, RuleML 0.7 DTDs have been developed into RuleML 0.8 DTDs without affecting the above semantics. The new markup uses XML in RDF's 'explicit role-markup' style, relativizing XML's positionality to places where RDF's Seq containers or DAML+OIL lists are used


RuleML DTDs: release 0.8

<fact> <_head> <atom> <_opr><rel href="http://www.eeebizz.com/rdf-sch#Approval"/></_opr> <ind href="http://www.eeebico.org/docs/joint-mission.html"/> <ind>full</ind> </atom> </_head></fact>

RDF-like triple:

Advantage of roles is object-centered modeling: If some item is to be added to a fact or atom, then this is easy to attach, RDF-like

Item insertion, XML-like, into child sequence would cause all subsequent children to assume a new position (problem for XSLT)

roles

Concrete markup:


Negation Handling in RuleML

IF isPresent(?someCar)

ANDNOT isAssignedToRentalOrder(?someCar)

ANDNOT requiresService(?someCar)

AND…

THEN isAvailable(?someCar)

The Closed-World Negation NOT

UML Diagram:


Negation Handling in RuleML (Cont’d)

<_body> <and> <atom> <_opr><rel>isPresent</rel></_opr> <var>Car</var> </atom> <not> <atom> <_opr><rel>isAssignedToRentalOrder</rel></_opr> <var>Car</var> </atom> </not> . . . </and> </_body>

NOT in RuleML: an example


Negation Handling in RuleML (Cont’d)

NOT = non-truth (or failure) NEG = falsity

Default rules:

IF NOT hasApproval(?URL)THEN NEG

isOfficialDocument(?URL)

Open-World Reasoning with NOT and NEG


Priorities/Evidences in RuleML

Quantitative Priorities: Static numerical salience Dynamically calculated priorities

Qualitative PrioritiesUsing the Overrides Fact over rule labels


Priorities/Evidences in RuleML (Cont’d)

<imp> <_rlab><ind>beginners</ind></_rlab> <_spriority><ind>0.75</ind></_spriority>

<_head> ....... </_head> <_body>

<atom> <_opr><rel>customerUnder25</rel></_opr>

<var>customer</var> </atom>

</_body></imp>

Static Salience Example:


Priorities/Evidences in RuleML (Cont’d)

Dynamic Salience Example:

<imp> <_rlab><ind>beginners</ind></_rlab> <_dpriority><rel>calculateSalience</rel></_dpriority>

<_head> ....... </_head> <_body>

<atom><_opr><rel>customerUnder25</rel></_opr>

<var>customer</var> </atom>

</_body></imp>


RuleML Implementations via XSLT & Rule

Engines

XSLT: is a Rule Based Language is a Transformation Tool for RuleML Knowledge

Bases (KB): Using Style sheets to exchange RuleML KBs

Implementations: With GEDCOM, Mike Dean created the first

operational RuleML (0.7) rulebase, where rules on family relationships (child, spouse, etc.) are run via XSLT translators to the XSB, JESS, and N3/cwm engines

With Mandarax RuleML, Jens Dietrich has implemented the first complete input-processing-output environment for RuleML (0.8)


A Semantic Web Scenario in Insurance

A teenager after her first car accident:

how much will my premiums increase and how does this accident affect my insurance policy?

Principle: Uniform RuleML representation of derivation rules and underlying document metadata


A Semantic Web Scenario in Insurance (Cont’d)

<imp> <_rlab><ind>beginners</ind></_rlab>

<_spriority><ind>0.75</ind></_spriority>

<_head> <atom> <_opr><rel>calculatePremium</rel></_opr> <var>customer</var> <ind>40</ind> </atom>

</_head> <_body><and>

<atom> <_opr><rel>InsurancePolicy</rel></_opr>

<var>customer</var><var>insurance</var>

</atom> <atom>

<_opr><rel>lifeEvent</rel></_opr> <var>customer</var><ind>accident</

ind><var>report</var> </atom>

<atom> <_opr><rel>customerUnder25</rel></_opr>

<var>customer</var>

</atom></and> </_body>

</imp>



<imp> <_rlab><ind>family</ind></_rlab>

<_spriority><ind>0.9</ind></_spriority>

<_head> <atom> <_opr><rel>calculatePremium</rel></_opr> <var>customer</var><ind>0</ind> </atom>

</_head> <_body><and>

<atom> <_opr><rel>InsurancePolicy</rel></_opr>

<var>customer</var><var>insurance</var>

</atom> <atom>

<_opr><rel>lifeEvent</rel></_opr> <var>customer</var><ind>accident</

ind><var>report</var> </atom>

<atom> <_opr><rel>FamilyAutoPlan</rel></_opr>

<var>customer</var><var>familyauto</var>

</atom></and> </_body>

</imp>



Formalizing the relevant facts

Olivia is under 25

Olivia has an insurance policy and this document has link .../IMA-0835:

Olivia is in a family auto plan and this document has link .../FMA-0142: Olivia's accident report is available at TrafficReport.biz : <fact>

<_head><atom> <_opr> <rel>FamilyAutoPlan</rel></_opr> <ind>Olivia</ind> <ind href = “http://www.BostonInsurance.com/policy/FMA-0142” /></atom>

</_head></fact>

<fact> <_head>

<atom> <_opr><rel>LifeEvent</rel></_opr> <ind>Olivia</ind> <ind>accident</ind> <ind href = “http://www.TrafficReport.biz/MA/report0712” /></atom>

</_head></fact>

<fact> <_head>

<atom> <_opr><rel>InsurancePolicy</rel></_opr> <ind>Olivia</ind> <ind href = “http://www.BostonInsurance.com/policy/IMA-0835” /></atom>

</_head></fact>

<fact> <_head>

<atom> <_opr><rel>customerUnder25</rel></_opr> <ind>Olivia</ind></atom>

</_head></fact>



Conflict Resolution: using Meta facts (a la BRML)

<imp> <_rlab>

<ind>beginners</ind></_rlab>

. . .</imp>

<imp><_rlab>

<ind>family</ind></_rlab>

. . .</imp>

<fact> <_head>

<atom>

<_opr><rel>Overrides</rel></_opr> <ind>family</ind> <ind>beginners</ind></atom>

</_head></fact>



Conflict Resolution: using Meta facts (a la BRML)

<imp> <_rlab>

<ind>beginners</ind></_rlab>

. . .</imp>

<imp> <_rlab>

<ind>family</ind></_rlab>

. . .</imp>

<fact> <_head>

<atom>

<_opr><rel>Overrides</rel></_opr> <ind>family</ind> <ind>beginners</ind></atom>

</_head></fact>


Conclusions

RuleML rulebases embedded in, or referring to, XML and RDF documents.

Using RuleML in other XML namespaces (like for, say, MathML), incl.

RDF(S) namespaces (ruleml: namespace prefix)

RuleML rule premises could be DAML+OIL concept descriptions orUML object descriptions

RuleML derivation rules integrated with reaction rules, e.g. Situated LP or ECA Framework (see SpiritSoft Demo)

RuleML and RDF Query talking about a joint Rule and Query effort as part of W3C Ontology Working Group

Ontologies:The big Picture and future RuleML Versions


Conclusions (Cont’d)

IJCAI 01: two tutorials, a workshop and an invited talk,August 2001, Seattle, USA

ICEC 01 (International Conference on Ecommerce) panel,November 2001, Vienna, Austria

Dagstuhl seminar on Rule Markup Techniques: February 2002, Schloss Dagstuhl, Germany

Some RuleML events

Documents

The RuleML Initiative, July/August 2001 International Semantic Web Working Symposium SWWS-01 30 July - 1 August 2001, Stanford, California, USA Harold