34
Ontology Design Pa/ern-driven Linked Data Publishing Adila Krisnadhi Data Seman1cs Lab (a.k.a. DaSeLab) Wright State University, Dayton, OH E-mail: [email protected] GitHub: krisnadhi 2016 ESIP Summer Mee1ng, Durham, NC

Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

OntologyDesignPa/ern-drivenLinkedDataPublishing

AdilaKrisnadhiDataSeman1csLab(a.k.a.DaSeLab)WrightStateUniversity,Dayton,OHE-mail:[email protected]:krisnadhi

2016ESIPSummerMee1ng,Durham,NC

Page 2: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Thistalkisabout…

Realizinginteroperabilitywithoutsacrificing(seman1c)heterogeneity.

2

Page 3: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Seman1cTechnology(again!)

•  Atleastmen1oned/introducedin…–  Bo[s,Fredericks,Gayanilo,Rueda.“BuildingSeman1candSyntac1cInteroperabilityIntoEnviroSensingSystems”(Tuesdaya_ernoon)

–  Narock.“OntologiesandtheSeman1cWeb-AnIntroduc1onforNon-Experts”(LateWednesdaya_ernoon)

3

Page 4: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Seman1cWebis…

h[ps://www.w3.org/2007/03/layerCake.png

“O_enseen,thoughnotallarerealized”

W3CSeman1cWebAc1vity(un1lendof2013)

W3CDataAc1vity(2014onward)•  WGonDataontheWebBest

Prac1ces•  WGonRDFDataShapes•  WGonSpa1alDataonthe

Web(JointwithOGC)•  SIGonHealthCareandLife

Sciences

4

Page 5: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Oralterna1vely…

Seman1cWeb

LinkedData

Vocabulary,Ontology

Inferencing,Querying,etc.

5

Page 6: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

LINKEDDATAPUBLISHING

6

Page 7: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

LinkedDataInaNutshell•  UsegraphdatamodelbasedonRDF.•  RDFgraphisasetofRDFtriples.•  RDFtripleconsistsof:–  Subject:URI,anonymousresource–  Predicate:URI– Object:URI,literal,anonymousresource.

•  Serializa1onformat:XML,Turtle,Ntriple,JSON-LD.•  Atriplecanexpressalinkingbetweenpiecesofdata.

•  Simplicityleadstopopularity.•  SeealsoCarlosRueda’sslidesonhowtotriplifytabular/rela1onaldata.

7

Page 8: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

LinkedDataGraph(of2Repos)

8

Page 9: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

9

StateofLinkedData

Page 10: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Howdoyoupublish(linked)data

10

•  LinkedDataPrinciples:– UseWebiden1fiers:HTTPURI/IRI–  EnsurethatURIsareWeb-resolvablesohumanANDmachinecanobtainfurtherinforma1onaboutthethingsURIsrepresented.•  Machine-processabledescrip1onàRDFgraph/triples.

– Asmuchaspossiblelinktodatafromotherpar1es.•  Inprac1ce,youneedtodecidehowto:–  Preparevocabularytodescribe/linkyourdata– MintURIsforyourdataandvocabulary

•  Incl.min1ngresolvableURIsforthevocabularytermsifnecessary.

–  SetupinfrastructuretoservethedataasLinkedData.

Page 11: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

ShouldImintURIforX?•  Google(2012):“Things,notstrings”•  IfXisinstancedata:

–  Do,ifXcomesfromyourownlocaldatabase/source.–  Don’t(i.e.,reuseexis1ngone),ifXoriginatesfromexternalsourceyou

don’tmaintain.•  IfXisavocabularyterm:

–  Do,ifthere’snoknownURIforXoryouwanttoassertyourowndefini1onforX(becauseitdoesnotexist,oryoudisliketheexis1ngone).•  Unlessthecurrentmaintainerofdefini1onofXagreeswithyour(new)defini1on.

–  Don’t,ifyoulikeexis1ngdefnanditfitsyourcurrentANDfutureneeds.•  Inanycase,ifyouDOdecidetomintanewURIforX,you’reresponsibleto

maintainit.èURIsmustbepersistent!•  URIsshouldpreferablybeopaqueèmachinesshouldnotparseorreadinto

URItoinferanythingaboutthereferencedresource;inferfromthedescrip1onofthedatainthegraph(theRDFtriples).

11

Page 12: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Otherthingstoconsider…•  HashURIvs.SlashURI– HashURI,e.g.:h[p://www.w3.org/ns/prov#wasAssociatedWith

–  SlashURI,e.g.:h[p://data.rvdata.us/id/award/100044•  Mayinvolvea303Redirect

–  seeh[ps://www.w3.org/TR/cooluris/andh[ps://www.w3.org/wiki/HashVsSlash

–  IpersonallyliketousehashURIforvocabularyterms,andslashURIfordatainstances

•  Namingconven1onforURIs–  CamelCase-ing?– Useof‘-’(dash)and/or‘_’(underscore),etc.

12

Page 13: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

EnsuringWeb-resolvabilityinaLinkedDataway

•  EverylookupofaURIshouldreturnsomething.•  Ifahuman-readabledescrip1onisrequested:– Usuallyindicatedbycontent-typeheadertext/html–  ReturnHTMLpage.

•  Ifamachine-readabledescrip1onisrequested:–  Indicatedbycontent-typeheader:application/rdf+xml,application/json,text/turtle,etc.

–  Returntheappropriateserializa1onformat.•  EasingtheURIpersistence:usepermanentredirec1onthroughPURLservice(seeh[p://www.purlz.org,h[ps://w3id.org/)

13

Page 14: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

VOCABULARYPREPARATION

14

Page 15: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

VocabularyandOntology

•  Ontology=formalizedvocabulary–  Formally,ontology=setoflogicalstatements(axioms)involvingthevocabularyterms.

–  Standardizedontologylanguages:RDFS,OWL–  Rule-basedlanguagesuchasRIFandSWRLcanalsobeused,thoughmorerarely.

•  Whyontologiesarevaluable(Janowicz,2016)?–  Improvediscoverabilityofyourowndata(asopposedtosimplekeywordsearch)

–  Cornerstoneofdatapublica1onandmanagingstrategies–  Improvedatareproducibility(throughprovenanceinforma1on)–  Easecross-repositoryknowledgeexplora1on(follow-your-nosebrowsing)

–  Easethedetec1onofinconsistencyinthedata.–  Enabledataintegra1on

15

Page 16: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Misconcep1onsaboutOntology

•  Misconcep1on#1:Thepurposeofontologyistoagreeonwhatthetermmeans.–  Correc1on:Itspurposeistomakeintendedmeaningexplicit.

•  Misconcep1on#2:Commonupper-leveland(large,overarching)domainontologiescouldsolvethemessinessofLinkedDataworld.–  Correc1on:differentandconflic1ngperspec1vesarenaturalintheopen,sothereisnowaytoforceeveryonetousethesameclassesandproper1es.

•  Misconcep1on#3:Ontologyconstrainsthewaythevocabularytermsareused.–  Correc1on:Ontologyemploysopen-worldassump1onandinferen1alseman1cs,

–  e.g.,specifyinga(global)domainrestric1onofapropertydoesnotconstrainthepropertyusage,insteaditaddsmoreinferences.

16

Page 17: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Wheretofindontologies/vocabularies?

•  LOV(LinkedOpenVocabulary)site-h[p://lov.okfn.org/•  W3Chostsseveralprominentontologies/vocabularies:–  Seeh[p://lov.okfn.org/dataset/lov/agents/W3C

•  ESIPrepositories:–  h[p://cor.esipfed.org/ont#/–  h[p://seman1cportal.esipfed.org/ontologies

•  OBOFoundry-h[p://www.obofoundry.org/•  ODPPortal-h[p://ontologydesignpa[erns.org/•  ODPPublicCatalog-h[p://www.gong.manchester.ac.uk/odp/html/

•  NCBOBioportal-h[p://bioportal.bioontology.org/

17

Page 18: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Reuseornot?•  Choosingappropriateontologiesessen1allydependsonwhatyouwanttodowiththem.–  Yourusecase:discovery?integra1on?Both?anythingelse?–  DoesontologyXdefinesthetermsyouneed?Doyoulike/agreewiththetermdefini1ons?IsXsufficientlyextendible

–  Ifyourneedscanonlybesa1sfiedbymul1pleontologies,doesusingthemtogetherleadtopoten1alproblems?

•  “Ihavebeentoldtoreuseotherontologies”=>Yes,butdon’tdoitatanearlystage!Startfirstwithprovidingyourowndefini1on;thenalignwithexis1ngontologieslater.–  mayleadtoconfusion(e.g.,FOAF,Organiza1ononto,vCard,orSchema.org?)andrestrictcrea1vity

– Mayleadtoendlessdiscussiononterms(nottomen1on:transla1ons)

18Source:OscarCorcho,2014

Page 19: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Ifanontologyneedstobedeveloped…...

•  Principle#1:Small>>>Large.–  Smallnessusuallyimpliessimplicity

•  Principle#2:Modular>>>monolithic.–  Easiertouseasbuildingblocks.–  Highlyextendibile–  Easilyunderstandable

•  Principle#3:Beawareofmul1pleperspec1ves.Strikeabalancebetweenfosteringinteroperabilityvs.allowingseman1cheterogeneity.–  e.g.,streetisaconnec1onbetweentwoplaces,butalsoasepara1onthatcutsahabitatintopieces.

•  Principle#4:Addhuman-readableannota1ons–  Improveunderstandability.

19

Page 20: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

OntologyDesignPa[ern(ODP)•  Isagoodcandidatew.r.tearlierprinciples•  ODP:reusablesolu1onofarecurrentmodelingproblem•  ContentODPs(akaknowledgepa[erns):ODPcorrespondingtoacoreno1oninapar1culardomain.–  Coverawiderangeofdomainsorapplica1onareas.–  Beextensibletoallowaddi1onaldetails;minimalontologycommitmentsfosteringreuse.

–  Beself-containedtoadegreewheretheycanbeusedontheirown.

–  Supportsmul1plegranulari1es.–  Provideanaxioma1za1onbeyondmeresurfaceseman1cs.–  Havevarioushookstowell-knownontologies/pa[erns.

20

Page 21: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

ExampleODP

21

VariantofSeman1ctrajectorypa[ern(Hu,etal.,2013).Axioma1za1onisalsoimportantpartofthepa[ern,butnotdisplayedhere.ConsulttheOWLencodingath[p://w3id.org/daselab/onto/trajectory

Page 22: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

ExampleODP(contd.)

22

•  DataprovidersA,B,andC,eachwiththeirownlocalontologies,butuseseman1ctrajectorypa[ernasacorecomponent.

•  A:dataabout(pedestrian)humanmobilitycapturedusingsmartphones,othermobiledevices,andsocialmedia.

•  B:dataaboutcars,buses,taxis,trucks,andsoforth.•  C:sparseGPS-basedwildlifetrackingdatafromCalifornianmountainlions.•  Federatedqueryexample:detectspotswherewildlifecrosseshighwaysor

entershumanse[lements.

Page 23: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

CruiseatR2R

23

Page 24: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

CruiseatBCO-DMO

24

Page 25: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Mynot-so-well-designedCruisepa[ern

25

Page 26: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Nextsteps

•  Fillinthelogicalaxioma1za1onofthepa[ern.– Useontologyeditors,e.g.,Protégé

•  Preparehuman-readableHTMLdocumenta1on.– E.g.,useLODE,Parrot,etc.

•  Makeboththepa[ernandthedocumenta1onavailableonlineaccordingthepa[ernURI(mayneedtosetupcontentnego1a1on)

•  Startpopulatethepa[ernwithdata(virtualorwarehousing-style).

26

Page 27: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

PUBLISHINGAGAINSTTHEPATTERNS

27

Page 28: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Localschemastopa[ernmapping

•  Mappingscanbeexpressedasrules/SPARQLConstructqueries/OWLaxioms[livedemorunningSPARQLqueriesontheR2RandBCO-DMOSPARQLendpoints]

•  R2R:–  gl:Cruise(x)->my:Cruise(x)–  gl:isUndertakenBy(x,y)->my:isUndertakenBy(x,y)–  r2r:hasAward(x,y)->my:fundedBy(x,y)–  etc.

•  BCO-DMO:–  odo:Cruise(x)->my:Cruise(x)–  odo:ofPla}orm(x,y)->my:isUndertakenBy(x,y)–  odo:Cruise(x),prov:associatedWith(x,y),odo:Project(y),odo:hasAward(y,z),odo:GrantAward(z)->my:Cruise(x),my:fundedBy(x,z),my:Award(z)

–  etc.

28

Page 29: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Interoperabilitythroughthepa[ern

•  Wecanmakedataavailableaccordingtothepa[ern.– Possibleevenwithoutphysicallypersistentlyhousingthedata.

– Mappingrulesareneeded(expressibleinSPARQL).•  R2RandBCO-DMOdonothavetoannotatetheirdatausingvocabularytermsinthepa[erndirectly.

•  Federatedquerycanalsobeposedinanyofthetworepositories’endpoints,assumingthecorrespondingrepositorycanreadthemapping.

29

Page 30: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Conclusion•  ODPscanactasinteroperabilitybridge,orasaglue,withoutsacrificingthelocalheterogeneityfromeachdatasource.

•  Thereisnoneedtoforceeveryonetousethesameclassandproper1es,asonecanmap/alignlocalschemas/datamodelstotheODPs.– HelpedbythefactthatODPsaresmallandmodular.

•  ODPsopenawaytopublishLinkedDatamorecheaplysincethecostlyendeavorofdevelopingoverarchingupperlevelanddomainontologiescanbeavoided.

30

Page 31: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

DataSeman1csLab

PascalHitzlerProfessor,LabCo-Director

MichelleCheathamAsst.Professor,LabCo-Director

OtherMembers:•  Postdoc:AdilaKrisnadhi•  8PhDStudents•  AfewMaster’sstudentsandvisi1ng

researchers

Web:h[p://www.daselab.orgTwi[er:@DaSeLabFB:h[ps://www.facebook.com/daselab

31

Page 32: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

References1)  O.Corcho,“OntologyEngineeringforandbythemasses:are

wealreadythere?”.KeynoteTalkatEKAW2014.2)  P.Hitzler,A.Gangemi,K.Janowicz,A.Krisnadhi,V.Presu~

(eds),OntologyEngineeringwithOntologyDesignPa7erns:Founda<onsandApplica<ons.IOSPress,2016.InPress.

3)  K.Janowicz,“ModelingOntologyDesignPa[ernswithDomainExperts–AViewFromtheTrenches”.In:(2)

4)  K.Janowicz,A.Gangemi,P.Hitzler,A.Krisnadhi,V.Presu~,“Introduc1on:OntologyDesignPa[ernsinaNutshell”.In:(2)

5)  Y.Hu,K.Janowicz,D.Carral,S.Scheider,W.Kuhn,G.Berg-Cross,P.Hitzler,M.Dean,andD.Kolas.Ageo-ontologydesignpa[ernforseman1ctrajectories.InSpa1alInforma1onTheory,pages438–456.Springer,2013.

32

Page 33: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Acknowledgement

33

AdilaKrisnadhiissupportedbytheNa1onalScienceFounda1onundertheaward1440202“EarthCubeBuildingBlocks:Collabora1veProposal:GeoLink-LeveragingSeman1csandLinkedDataforDataSharingandDiscoveryintheGeosciences.”

Page 34: Ontology Design Paern-driven Linked Data Publishing · Linked Data world. – Correc1on: different and conflic1ng perspec1ves are natural in the open, so there is no way to force

Thankyou!

Specialthanksto:AdamShepherd(BCO-DMO)&BobArko(R2R)

34