Owl2 rl

OWL2 RL (OWLIM)Zaenal Akbar

April 10, 2014

Copyright 2014 STI INNSBRUCK www.sti-innsbruck.at

Outline

OWL2 RLOWLIMThe Weaver

Rule Based SystemRDF Reasoner

References

www.sti-innsbruck.at 2/21

OWL - Web Ontology Language

OWL is a semantic fragment of FOL 1

OWL exists in different flavorsOWL Lite ⊆ OWL DL ⊆ OWL FullOWL2 EL, OWL2 RL, OWL2 QL ⊆ OWL2 DL ⊆ OWL2 Full

OWL 2 is a very expressive language, hard to implementand work withOWL 2 specifies profiles/fragments/sub-languages→trades some expressive power for the efficiency ofreasoningOWL 2 RL:

Designed to be amenable to implementation usingrule-based technologiesPresenting a partial axiomatization of the OWL 2RDF-Based Semantics

1H. Sack, Semantic Web Technologies - OWL 2, 2012,http://www.slideshare.net/lysander07/10-semantic-web-technologies-owl-2


http://www.slideshare.net/lysander07/10-semantic-web-technologies-owl-2

OWLIM

OWLIM is a high-performance semantic repositoryImplemented in Java and packaged as a Storage andInference Layer (SAIL) for the Sesame RDF database [2]


OWLIM

Implemented on top of the TRREE (Triple Reasoning andRule Entailment Engine)Performs reasoning based on forward-chaining ofentailment rules over RDF triple patterns with variablesR-Entailment 2

Premises and conclusions are triple patternsVariables allowed in any positionInequality constraintsRules applied directly to RDF graph

2B. Bishop, S. Bojanov, Implementing OWL 2 RL and OWL 2 QL rule-sets for OWLIM, 2011www.sti-innsbruck.at 5/21

OWLIM - Semantics

OWLIM offers several predefined semantics by way ofstandard rule-sets

1. empty: no reasoning, i.e. operates as a plain RDF store2. rdfs: supports standard RDFS semantics3. owl-horst: OWL dialect close to OWL Horst4. owl-max: a combination of most of OWL-Lite with RDFS5. owl2-ql: Conformant OWL2 QL profile6. owl2-rl: Fully conformant OWL2 RL profile, except for

D-Entailment

Can be configured to use custom rule sets with semanticsbetter tuned to the particular domain


OWLIM - Reasoner (1)

Prefices {

rdfs : http://www.w3.org/2000/01/rdf-schema#

}

Axioms {

<rdf:type> <rdf:type> <rdf:Property>

<rdf:subject> <rdf:type> <rdf:Property>

}

Rules {

Id: <rule_name>

<premises> <optional_constraints>

-------------------------------

<consequences> <optional_constraints>

}


OWLIM - Reasoner (2)

Id: prp_trp

p <rdf:type> <owl:TransitiveProperty>

x p y

y p z

-------------------------------

x p z

Id: prp_key_1

c <owl:hasKey> u

x <rdf:type> c [Constraint x != y]

y <rdf:type> c [Cut]

x u y [Context <onto:_sameKey>]

-------------------------------

x <owl:sameAs> y


The Weaver

Bottleneck: serialization/deserializationHow to bring the reasoner closer to the data?


RDF Model (1)

@prefix owl: <http://www.w3.org/2002/07/owl#> .

@prefix rdf:

<http://www.w3.org/1999/02/22-rdf-syntax-ns#> .

@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .

@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .

@prefix schema: <http://schema.org/> .

@prefix oc: <http://sti2.at/oc/> .


RDF Model (2)

schema:Event a rdfs:Class;

rdfs:label "Event"@en;

rdfs:subClassOf schema:Thing;

rdfs:isDefinedBy <http://schema.org/Event> .

schema:description a rdf:Property; ...

schema:name a rdf:Property; ...

schema:startDate a rdf:Property; ...

schema:endDate a rdf:Property; ...

schema:location a rdf:Property;

rdfs:domain [ ... owl:unionOf(schema:Event ...) ];

rdfs:range [ ... owl:unionOf(schema:Place ...) ];

schema:Place a rdfs:Class; ...


RDF Model (3)

oc:Channel a rdfs:Class;

rdfs:label "Channel"@en;

rdfs:isDefinedBy <http://sti2.at/oc/Channel> .

oc:Twitter a rdfs:Class;

rdfs:label "Channel Twitter"@en;

rdfs:subClassOf oc:Channel;

rdfs:isDefinedBy <http://sti2.at/oc/Twitter> .

oc:Blog a rdfs:Class;

rdfs:label "Channel Blog"@en;

rdfs:subClassOf oc:Channel;

rdfs:isDefinedBy <http://sti2.at/oc/Blog> .


RDF Model (4)

oc:publish a rdf:Property;

rdfs:label "Publish"@en;

rdfs:comment "Publish a Content to a Channel."@en;

rdfs:domain [ a owl:Class;

owl:unionOf (schema:Event schema:BlogPosting) ];

rdfs:range [ a owl:Class;

owl:unionOf (oc:Twitter oc:Blog) ];

rdfs:isDefinedBy <http://schema.org/Event>;

rdfs:isDefinedBy <http://schema.org/BlogPosting>;

.


Rules (1)

Prefices {

schema : http://schema.org/

oc : http://sti2.at/oc/

}

Axioms {

...

<schema:Event> <rdf:type> <rdfs:Class>

<schema:Event> <rdfs:subClassOf> <rdfs:Resource>

<oc:Twitter> <rdf:type> <rdfs:Class>

<oc:Twitter> <rdfs:subClassOf> <rdfs:Resource>

...

<oc:publish> <rdf:type> <rdf:Property>

<oc:publish> <rdfs:domain> <rdfs:Class>

<oc:publish> <rdfs:range> <rdfs:Class>

}


Rules (2)

Rules

{

Id: pub_event

x <rdf:type> <schema:Event>

-------------------------------

x <oc:publish> <oc:Twitter>

x <oc:publish> <oc:Blog>

Id: pub_post

x <rdf:type> <schema:BlogPosting>

-------------------------------

x <oc:publish> <oc:Blog>

...

}


Instances (1)

oc:3W04 a schema:Place ;

schema:description "Room 3W04" .

oc:M1 a schema:Event ;

schema:name "OC Meeting" ;

schema:location oc:3W04 ;

schema:startDate "2014-04-09T14:30" ;

schema:endDate "2014-04-09T15:30" .

Test> sparql prefix oc:...

insert data { oc:M1 a schema:Event; ... }.

Executing update...

Update executed in 431 ms


Instances (2)

Test> sparql

select * where {?s <http://sti2.at/oc/publish> ?o}.

Evaluating query...

+-------------------------+---------------------------+

| s | o |

+-------------------------+---------------------------+

| oc:M1 | oc:Twitter |

| oc:M1 | oc:Blog |

+-------------------------+---------------------------+

2 result(s) (43 ms)

Test>


Comparison (1)

OWLIM’s Custom Rule-Sets:Specify the custom rule-set manually via the configurationparameter (-Druleset=/opt/owlim-se-5.3/oc.pie)The rule-set file is processed to create Java source code,then compiled using JDK, used by TRREE during inference

$ tail -f logs/catalina.out

OwlimSchemaRepository: version: 5.3, revision: 6156

Product: OWLIM_SE

Licensee: STI_PlanetData

...

Current file: /opt/owlim-se-5.3/oc.pie

Compiled: ’/opt/owlim-se-5.3/oc.pie’


Comparison (2)

OWLIM’s Reasoner:⊕ Compact (no more objects instantiation), Centralized (nomore scattered servers) Less flexible to maintain the Rules Rules filtering still questionableIF (event.location == "Innsbruck") THEN ...

Drools (+Guvnor) Rules Engine:⊕ More flexible on Rules editing, i.e. GUI, DSL, filtering⊕ KB sessions can be created/destroyed dynamically Complex, requires objects instantiation


Summary

OWLIM can simplify the complexity of the Weaver, andmore than enough for the most use casesShould we use OWLIM’s Reasoner instead?


References

1. OWLIM, http://owlim.ontotext.com2. OpenRDF Sesame, http://www.openrdf.org3. OWL 2 Primer, http://www.w3.org/TR/owl2-primer/4. OWL 2 Profiles, http://www.w3.org/TR/owl2-profiles/5. Schema.org, http://schema.org/6. Drools, http://www.jboss.org/drools/


http://owlim.ontotext.com

http://www.openrdf.org

http://www.w3.org/TR/owl2-primer/

http://www.w3.org/TR/owl2-profiles/

http://schema.org/

http://www.jboss.org/drools/

Owl2 rl

Presentations & Public Speaking