The Ontological Foundation of REA Enterprise Information Systems.docx
《The Ontological Foundation of REA Enterprise Information Systems.docx》由会员分享,可在线阅读,更多相关《The Ontological Foundation of REA Enterprise Information Systems.docx(31页珍藏版)》请在冰豆网上搜索。
TheOntologicalFoundationofREAEnterpriseInformationSystems
TheOntologicalFoundationofREAEnterpriseInformationSystems
November1999,March2000,August2000
GuidoL.Geerts
AssistantProfessorofAccounting&MIS
TheUniversityofDelaware
Newark,DE19711
EMAIL:
geertsg@be.udel.edu;TEL:
302-831-6413
&
WilliamE.McCarthy
ProfessorofAccountingMichiganStateUniversityEastLansing,MI48864USA
EMAIL:
mccarth4@msu.edu;TEL:
517-432-2913
CommentswelcomedThispaperhasbenefitedfromworkshopcommentsreceivedattheUniversityofFlorida,theUniversityofDelaware,ArizonaStateUniversity,theUniversityofSanDiego,theUniversityofKansas,VirginiaTechUniversity,theUniversityofSouthFlorida,theUniversityofMinnesota,theUniversityofWisconsin,andthe2000AAANationalMeeting.ThesuggestionsofCherylDunnandJulieSmithDavidareespeciallyacknowledged.TheontologicalconstructsandtheirdefinitionshavealsobenefitedtremendouslyfromcritiquesandcommentsoftheebXMLBusinessProcessTeam,especiallyPaulLevine(Telecordia),KarstenRiemer(Sun),andJimClark(Edifecs).OurmostsignificantacknowledgementgoestoBobHaugen(LogisticalSoftware)whoseinsightfulcommentariesandcritiqueshavecausedmultiplechangestotheontology.
TheOntologicalFoundationofREAEnterpriseInformationSystems
Authorsandaffiliations
GuidoGeerts,TheUniversityofDelaware
WilliamE.McCarthy,MichiganStateUniversity
ABSTRACT:
Philosophershavestudiedontologiesforcenturiesintheirsearchforasystematicexplanationofexistence:
“Whatkindofthingsexist?
”Recently,ontologieshaveemergedasamajorresearchtopicinthefieldsofartificialintelligenceandknowledgemanagementwheretheyaddressthecontentissue:
“Whatkindofthingsshouldwerepresent?
”Theanswertothatquestiondifferswiththescopeoftheontology.Ontologiesthataresubject-independentarecalledupper-levelontologies,andtheyattempttodefineconceptsthataresharedbyalldomains,suchastimeandspace.Domainontologies,ontheotherhand,attempttodefinethethingsthatarerelevanttoaspecificapplicationdomain.BothtypesofontologiesarebecomingincreasinglyimportantintheeraoftheInternetwhereconsistentandmachine-readablesemanticdefinitionsofeconomicphenomenabecomethelanguageofe-commerce.Inthispaper,weproposetheconceptualaccountingframeworkoftheResource-Event-Agent(REA)modelofMcCarthy(1982)asanenterprisedomainontology,andwebuildupontheinitialontologyworkofGeertsandMcCarthy(2000)whichexploredREAwithrespecttotheontologicalcategorizationsofJohnSowa(1999).Becauseofitsconceptualmodelingheritage,REAalreadyresemblesanestablishedontologyinmanydeclarative(categories)andprocedural(axioms)respects,andwealsoproposeheretoextendformallythatframeworkboth
(1)verticallyintermsofentrepreneuriallogic(valuechains)andworkflowdetail,and
(2)horizontallyintermsoftypeandcommitmentimagesofenterpriseeconomicphenomena.Astrongemphasisthroughoutthepaperisgiventothemicroeconomicfoundationsofthecategorydefinitions.
I.INTRODUCTION
Philosophershavestudiedontologiesforcenturiesintheirsearchforasystematicexplanationofexistence:
“Whatkindofthingsexist?
”Recently,ontologieshaveemergedasamajorresearchtopicinInformationSystemswheretheyaddressthecontentissue:
“Whatkindofthingsshouldwerepresent?
”Theanswertothatquestiondifferswiththescopeoftheontology.Ontologiesthataresubject-independentarecalledupper-levelontologies,andtheyattempttodefineconceptsthataresharedbyalldomains,suchastimeandspace.Examplesofupper-levelontologiesareCYC(LenatandGuha1990),JohnSowa’sontology(Sowa1999),andtheWeberandWandontology(WandandWeber1990,Weber1997)basedontheworkofMarioBunge(1977,1979).Domainontologies,ontheotherhand,attempttodefinethethingsthatarerelevanttoaspecificapplicationdomain.Examplesofdomain-specificontologiesincludeaircampaignplanning(Valenteetal.1999),scheduling(SmithandBecker1997),andmedicine(Gangemietal.1998).Inthispaper,wegenerallyexploretheareaofenterpriseontologies(Foxetal.1993,Uscholdetal.1998),andwespecificallyproposeaconceptualaccountingframework–theResource-Event-Agent(REA)model(McCarthy1979,1982)--asanenterpriseontology.Becauseofitsconceptualmodelingheritage,REAalreadyresemblesanestablishedontology,andwealsoproposeheretoextendthatframeworkbothverticallyintermsofentrepreneuriallogicandworkflowdetailandhorizontallyintermsofphysical-abstractcharacterizationsofenterpriseeconomicphenomena.Beforewestartourexplorationweneedtoaddressanumberofbroaderissues:
“Whatisanontology?
”“Whystudyontologies?
”“Howtoconstructanontology?
”and“Whatisagoodontology?
”Answerstothesequestionshelpustoframetheresearchquestionaddressedinthispaper.
Whatisanontology?
ThemostwidelyaccepteddefinitionofontologyistheonegivenbyGruber(1993):
“anexplicitspecificationofconceptualization.”GruberusesthedescriptiongiveninGeneserethandNilsson(1987)tofurtherrefinethetermconceptualizationas“theobjects,concepts,andotherentitiesthatareassumedtoexistinsomeareaofinterestandtherelationshipsthatholdamongthem.”Thisdefinitionresemblesthetraditionaldescriptionofadatabaseconceptualschema;however,itdoesdifferinatleastthreeimportantways:
objective,scopeandcontent.First,theobjectiveofanontologyistorepresentaconceptualizationthatisshareable/reusableandwhereidiosyncrasiesofspecificapplicationsareignored.Second,thescopeofanontologyisallapplicationsinthedomain,notjustone.Andfinally,anontologycontainsknowledgespecificationswherethemeaningofthestructuresrepresentedisexplicitlyspecifiedandconstrainedandwheretherulestoinferfurtherknowledgeareexplicitlydefined.
Whystudyontologies?
Theinterestinontologieshasemergedinthecontextofthecurrentdistributed,heterogeneouscomputingenvironment--inparticulartheInternet--andinthefastgrowinginterestincomponent-basedsoftwareengineering.Thegeneralconsensusisthatontologiesareabletoimprovecommunication,sharingandreuse(UsholdandGruninger1996).Lackofanexplicitlyspecifiedconceptualizationoftenresultsinpoorcommunicationfrompeopletopeople,frompeopletocomputers,andespeciallyfromcomputerstocomputers.Forexample:
"Whatisthemeaningofthetermsaccount,businessprocessandmarket?
Interpretationofthesetermsisdifferentwithindepartments,acrossdepartments,acrossorganizations,andacrosscomputersystems.Animportantobjectiveofontologiesistomakethemeaningofconceptsexplicitinordertoimprovecommunication.Interoperability,thecommunicationbetweenseparatecomputersystems,hasbeengivenincreasedattentionwiththeemergenceofdistributedandheterogeneouscomputingenvironments,anddifferentapproacheshavebeenproposedinsupportofinteroperability.Inafirstapproach,knowledgeistranslatedintoacommonformatsuchasKIF–theKnowledgeInterchangeFormat(Fikesetal.1991)--andthisintermediateformatcanthenbeusedforknowledgesharingandreuse.AsecondapproachistheuseofanAgentCommunicationLanguage(ACL)thatisgroundedinanontology.AgoodexampleofanACLisKQML(Fininetal.1994).Agentsusetermsthatarepartoftheontologytocommunicatewiththeunderlyingimplementationbeingirrelevant.
Lackofreusabilityhasbeenwidelyrecognizedasamajorweaknessoftraditionalsystemsdevelopment.Reusecansubstantiallyreducetimeandcostofinformationsystemsdesign,implementationandmaintenance.Inrecentyearsmuchattentionisgiventothedesignofsoftwareapplicationswherepartsofexistingapplicationscanbereused.Thesepartsmightincludedesigns,knowledgestructures,orsoftwarecomponents.Theaccomplishmentofreusabilitylargelydependsonthesharingofasimilarconceptualization.
Manyusesofontologieshaveemergedwithimportantdifferencesinsophisticationandobjectives(Musen1992,UsholdandGruninger1996).Here,wedistinguishamongthreebroadcategoriesofontologyuse:
(1)asaknowledgedictionary,
(2)asasupportforconceptualdesign,and(3)inoperationaluse.
∙AKnowledgeDictionaryexplicitlyrecordsthemeaningofthedomainconcepts,therelationshipsbetweenconcepts,andtheconstraintsthatapplytoconcepts.Theexplicitlyrecordeddefinitionsimprovecommunication,integrationandconsistency.
∙ConceptualDesignSupport.Ontologiesofferimportantguidanceforconstructionofapplicationmodelsinaspecificdomain.AsUsholdandGruninger(1996)pointout,benefitsincludeabetteridentificationofrequirementsandincreasedreliability.
∙ForOperationalOntologies,theconcepts,relationshipsbetweenconceptsandconstraintsareexplicitlyrecorded,andthesethenbecomepartoftheapplicationsthemselves(Guarino,1998).Theexplicitrecordingoftheontologyasknowledgespecificationsenablestheiruseforinference.
Theseusesareneitherexhaustivenorexclusive.Forexample,conceptualdesignsupportbenefitsfromtheexistenceofaknowledgedictionary,andanexcellentexampleofoperationaluseofanontologyistheautomatedsupportofconceptualdesign.
Howtoconstructanontology?
Recognitionofthecomplexityofontologicalengineeringhasresultedinincreasedresearchonmethodologiesforontologyconstruction.Gomez-Perez(1998),forexampledistinguishesthefollowingphasesinontologyconstruction:
knowledgeacquisiti