DLP – An introduction

DennyVrandeˇci´c,PeterHaase,PascalHitzler,YorkSure,RudiStuder

InstituteAIFB

UniversityofKarlsruhe,Germany

dlp@aifb.uni-karlsruhe.de

Abstract

DLP–DescriptionLogicPrograms–isthenameforthecommonlanguagethatisabletointegrateknowledgebasesdescribedinDescriptionLogicwithLogicPrograms.Inthisintroduction,weofferaveryshortoverviewofDLP,themotivationforit,thebenefitsitoffersandhowtouseit.

1Introduction

Inthepastfewyears,theW3ChasestablishedstandardsforthefundamentalbuildingblocksoftheSemanticWeb.WithRDFacommondatamodelhasbeenprovided.AmajorstepwastheintroductionoftheontologydefinitionlanguagesRDFS–offeringonlyverybasicelementsfordescribingcon-ceptsandtheirrelationships–andthepowerfulWebOntol-ogyLanguageOWL,withthesublanguagesOWLDLandOWLLite,whicharebasedonDescriptionLogics[Baaderetal.,2003].ButOWLDL(andevenOWLLite)sufferfromaveryhighcomputationalcomplexity,andefficientreasonersabletodealwiththeminasufficientlyscalablewayarestilloneofthemainchallengesforthecommunity.

Asthesestandardsarestillverynew,quiteanumberofdifferentknowledgemodellinglanguagescontinuetoexist,mostlybasedonknowledgerepresentationparadigmsotherthan[DescriptionLogics.AnothermajorparadigmisF-LogicKiferetal.,1995;AngeleandLausen,2004],whichisstronglyrelatedtologicandobject-orientedprogramming,andallowsforthemodellingofknowledgebyimplicationrules,afeaturemissinginOWL.

2Motivation

Whenchoosingaknowledgerepresentationslanguagetoday,onehaseithertogoforanotyetefficientlyimplemented,butstandardizedlanguage,thuscommittingtowaituntilrea-sonableefficienttoolsareavailable,orforefficientsystemsbuildingonnon-standardizedlanguages,whicharealreadywellestablished.Forapplyingsemantictechnologies,thepractitionerisfacedwiththequestion,whichoftheabove—orother—paradigmsshouldbeadopted.Forperformancereasons,simplerparadigmsareoftenpreferable,butequallyimportantaretheinteroperabilitybetweendifferentsystemsandthefuturereusability.

ThisiswhereDLPentersthescene.KnowledgebasesinDLP–roughlyspeaking–liewithintheintersectionofDe-criptionLogicsandLogicProgramming.SoDLPprovidestheformallinkbetweenthetwomajorknowledgerepresen-tationparadigms,andservemultiplepurposesintheoreticalresearchandpracticalapplications.Followingthepathofin-creasingexpressivity,DLPdefinespreciselythepointofnoreturnbeforecommittingtooneoftheparadigms.Sincetheirfirstintroduction[Grosofetal.,2003],avarietyofresearchprojectshavebeguntoinvestigateandusethelanguage.

ThemajoradvantageofDLPisitsobviousfuturestabilitywithbothcurrentlyexistingmajortrendsforontologyrepre-sentation,descriptionlogics(OWLDL)andlogicprogram-ming(F-Logic).Bothhavebenefitsanddrawbacksandallowfordifferentusagescenarios,whichiswhytherearecurrentlyseveraleffortsbeingundertakenforjoiningbothparadigmsintoasingleframework.DLPoffersthisframework,head-ingformaximalflexibility,asitcaneasilybetranslatedfromoneparadigmtotheother.Wecandecidefreelytousethemostsuitableapproachbasedonthetaskathand,orevenmixbothparadigmsaswelike.WithDLPwegaintheabilitytousewell-testedandefficientimplementationswithamodernW3Cstandardizedlanguage.

DLPisnotyetanotherknowledgerepresentationlanguage,butratherabridgetobringtwoparadigmstogether.AssuchitimposescertainconstraintsonOWLDLinordertoguaranteethatallaxiomsstatedaretransformableinanefficientwaytoHornclauses,i.e.rulesinthesenseoftraditionallogicprogramming,butitdoesnotdefinenewsemanticsorsyntax.

3Benefits

Aswehaveseen,byusingDLPweoffermaximalflexibil-itytowardspossiblefuturedevelopmentsinknowledgerep-resentation.Thisinstantlybringsanothermajorbenefit:asthesemanticsremainthesamewithinDLP,wecanusetoolsfrombothworlds–thewell-equippedworldofF-LogicandtheheavilyresearchedworldofDL–inordertogenerate,evolve,present,workwithandreasonoverDLPknowledgebases.

ByprovidingsyntacticalconvertersweareabletojointhepowerofthenewestOWLtoolsandtheefficientandstablelogicprograminterpreterslikeOntoBroker,SWI-orXSB-Prolog.Efficientreasonerslikethese,existingalreadytoday,willprovidesufficientsupportinordertoworkwithOWL

immediately.Thepricewepayareafewlimitation,whichinpractisewillonlyveryrarelylimittheactualmodellingpossibilities,aswewillshowinthenextsection.

Ontheotherside,everyOWLDLPontologyisaOWLDLontologyaswell.DLPisapropersubsetofOWLDL,stayingtotallytruthfultothestandardisedsemantics–un-likeotherexistingproprietaryextensions,whichchangetheagreed-onsemanticsandthusbreaktheavailabletoolandhampercompatibilityandexchange.ThusweremainfaithfultosomeidentifiedbasicrequirementsfortheSemanticWeb,liketheOpenWorldAssumption,ClassicalNegationandtheNoUniqueNameAssumption.

Finally,theconstraintsimposedonDLPregardingthemoreexpressiveOWLDL,leadtoDLPenjoyingamuchbet-terdatacomplexity(polynomial)andcombinedcomplexity(exptime).Thisallowstoexpectmoreefficientandrespon-sivetoolsthanfullOWLDLreasoningwilleverbeabletoachieveduetothecomplexityofDLreasoningalgorithms.Notonlydoweachievetohaveanimmediatetoolsupport,butwecanalsostatethatreasoningoverDLPwillalwaysleadtoalowercomplexitythanreasoningoverDLcould.InthenextsectionwewilltakealookathowrestrictingtheconstraintsthatleadfromDLtoDLPreallyareinappliedontologies.

5LanguageDescription

OWLDLPisnoteasilydefined,butinordertofacilitatetheusageofOWLDLP,inthissectionadescriptionofaneasytousesublanguageofDLPisprovidedbylistingallthecon-structorsyoumayusefreelyinanOWLontologywithoutrunningtheriskofleavingDLP(takingforgrantedthatyousticktotheusualOWLDLconstraints).

Thisdoesn’tmeanthatotherconstructorsareforbiddeninOWLDLP.Itisjustthattheiruseunderliesfurtherconstraintswhichcan’tbedescribedinthisshortintroduction.

AllowedOWLconstructors:Class,Thing,subClass-Of,Property,subPropertyOf,domain,range,Individual,equivalentClass,equivalentProperty,sameAs,differentFrom,AllDifferent,ObjectProperty,DatatypeProperty,inverseOf,TransitiveProperty,SymmetricProperty,FunctionalProperty.InverseFunctionalProperty,intersectionOf

Asstated,thisisbynomeansadescriptionofthewholeDLPlanguage.Butitissupposedtohelptremendouslybymakingyoufeelsafewiththeusageoftheconstructors.

6FurtherReading

4Usage

ByrestrictingattentiontoDLP,wemakesignificantgainsinreasoningcomplexity,andgetaccesstoamuchbroadersup-portwithtools.Atthesametime,thesemanticexpressivitylostduetotherestrictiontoDLPhardlymattersinpractice:[Volz,2004]analysedthat99%oftheaxiomsinontologiestakenfromthedaml.orgrepositoryarewithintheDLPfrag-ment.

InordertouseDLP,thereisnoneedforspecialsoftware.StandardOWLeditorscanbeusedtocreatevalidDLPon-tologies.AsmostofthecommonusedDLconstructors,es-peciallythoseusedforlight-weightontologiesortaxonomieswithrelations,arewithinDLPanyway,DLPhasahead-startonuserscomparedtoalternativeapproachestogainhigherefficiencyinreasoning.

WeprovideaKAON2basedtoolthatconvertsDLPon-tologieswritteninOWL/XMLorRDF/XMLsyntaxtoLogicprogramsyntax,retainingthesemantics1,thusallowingtheuseofstandardout-of-the-boxLPreasonersforDLPreason-ing.FurtherweplantointegrateDLPintoourOWLevolutionframeworkevOWLution[Haaseetal.,2004]inordertopro-videthemeanstodevelopOWLontologieswiththeguaran-teetoremainwithinDLP.FinallyadaptingexistingontologyeditorsinordertohelptheusertoaccountforDLPauto-maticallyisplanned,thusenablingthestandardusertofullyleveragethepowerofDLP.DuetothesemanticentailmentofDLPinDL,thereisnoneedforrecreatingthewholerangeofontologymanagementtools,butonlysmalladjustments–ifany–ofexistingsoftwarewillallowtoreapfullbenefitoftheexistingsystemsandmethodologies.

ThebasicDLPlanguagewasdescribedinitiallyin[Grosofetal.,2003].[Volz,2004]providesamuchdeeperdefinitionandanalysisofDLP,especiallyevaluatingthescopeoftheidentifiedconstraintsandtheirpracticalmeaning.Aweb-sitewithpointerstofurtherresourcesisbeingmaintainedathttp://logic.aifb.uni-karlsruhe.de.Ques-tionsmaybesenttodlp@aifb.uni-karlsruhe.de.

References

[AngeleandLausen,2004]J¨urgenAngeleandGeorgLausen.OntologiesinF-logic.InSteffenStaabandRudiStuder,editors,HandbookonOntologies,pages29–50.Springer,2004.

[Baaderetal.,2003]FranzBaader,DiegoCalvanese,Deb-orahMcGuinness,DanieleNardi,andPeterPatel-Schneider,editors.TheDescriptionLogicHandbook:Theory,Implementation,andApplications.CambridgeUniversityPress,2003.

[Grosofetal.,2003]BenjaminGrosof,IanHorrocks,RaphaelVolz,andStefanDecker.Descriptionlogicprograms:Combininglogicprogramswithdescriptionlogics.InProc.ofWWW2003,Budapest,Hungary,May2003,pages48–57.ACM,2003.

[Haaseetal.,2004]PeterHaase,YorkSure,andDennyVrandecic.Ontologymanagementandevolution–survey,methodsandprototype.SEKTformaldeliverableD3.1.1,InstituteAIFB,UniversityofKarlsruhe,December2004.[Kiferetal.,1995]MichaelKifer,GeorgLausen,andJamesWu.Logicalfoundationsofobject-orientedandframe-basedlanguages.JournaloftheACM,42:741–843,1995.[Volz,2004]RaphaelVolz.WebOntologyReasoningwithLogicDatabases.PhDthesis,AIFB,UniversityofKarl-1

sruhe,2004.http://www.ubka.uni-karlsruhe.http://logic.aifb.uni-karlsruhe.de/dlpconvert

de/vvv/2004/wiwi/2/2.pdf.