Knowledge representation

Structure

Therepresentationofknowledgeisadescriptionofknowledge,orasetofconventionsonknowledge,adatastructureacceptabletocomputersfordescribingknowledge.Inasense,representationcanberegardedasasynthesisofdatastructureanditsprocessingmechanism:representation=datastructure+processingmechanism.Therefore,knowledgerepresentationinESisaseriesoftechnicalmeansthatcancompletethecomputerprocessingofexpertknowledgeinES.Thecommononesareproductionrules,semanticweb,framemethodandsoon.

Representsthesubject

Knowledgeisthecorrectconclusionobtainedbytheinformationreceiverthroughtherefiningandreasoningoftheinformation;itisthehuman'sunderstandingofthenaturalworld,humansociety,andthelawofthinkingandmovement.Recognitionandmasteryaretherecombinationandsystematicinformationcollectionofthehumanbrainthroughthinking.

InKR,themeaningofknowledgeisdifferentfromthemeaningofknowledgegenerallyknowntous.Itreferstoconcepts,eventsandprocessesexpressedinacertainstructuredway.Therefore,inKR,notallknowledgeindailylifecanbeembodied,butonlywhenthescopeandstructurearelimited,theknowledgethathasbeencodedandtransformedcanbecometheknowledgeinKR.TheknowledgeinKRgenerallyhasthefollowingcategories:

Conceptsabouttheobjectsofinterestintherealworld,thatis,usedtodescribetheabstractandsummarizedconceptsintherealworld.

Abouteventsthatoccurintherealworld,thebehaviorsandstatesofrelatedobjects,etc.,thatistosay,therearenotonlystaticconcepts,butalsodynamicinformation.

Knowledgeabouttheprocess,thatis,notonlythedescriptionofthecurrentstateandbehavior,butalsothedescriptionofthechangesinitsdevelopmentanditsrelatedconditions,causality,etc.

Meta-knowledge,thatis,knowledgeaboutknowledge,forexample,includingknowledgeabouttheuseofknowledge.

Method

Throughthejointeffortsofdomesticandforeignscholars,manyknowledgerepresentationmethodshavebeenstudiedindepth,andthemostusedknowledgerepresentationmethodsmainlyincludethefollowingknowledgerepresentationmethods.

(1)Logicalnotation

Logicalnotationexpressesthesubjectandobjectofactionsintheformofpredicates,whichisanarrativeknowledgerepresentationmethod.Usinglogicalformulas,peoplecandescribeobjects,properties,conditions,andrelationships.Itismainlyusedfortheproofofautomatictheorems.Logicalnotationismainlydividedintopropositionallogicandpredicatelogic.

Logicalrepresentationstudiestheimplicationrelationshipbetweenhypothesisandconclusion,thatis,thelawofreasoningwithlogicalmethods.Itcanbeseenasasimplifiedformofnaturallanguage.Becauseofitsaccuracyandunambiguousness,itiseasytounderstandandoperatebycomputers,andatthesametimeitissimilartonaturallanguage.

Propositionallogicisakindofmathematicallogic.Mathematicallogicusesformallanguage(logicalsymbollanguage)forprecise(withoutambiguity)descriptionandmathematicalresearch.Whatwearemostfamiliarwithistherepresentationofunknownsinmathematics.Example:Usepropositionallogictoexpressthefollowingknowledge:

Ifaisanevennumber,thena2isanevennumber.

Solution:Definethepropositionasfollows:P:aisanevennumber;Q:a2isanevennumber,then:theoriginalknowledgeisexpressedas:P→Q

ThepredicatelogicisequivalenttothefunctioninmathematicsExpress.Example:Usepredicatelogictoexpressknowledge:naturalnumbersareallintegersgreaterthanorequaltozero

Solution:definethepredicateasfollows:N(x):xisanaturalnumber;I(x):xisaninteger;GZ(x):xisanumbergreaterthanorequaltozero.Sotheoriginalknowledgeisexpressedas:(∀x)(N(x)(GZ(x)∧I(x)),∀(x)istheuniversalquantifier.

(2)Productionexpression

Productionexpression,alsoknownasruleexpression,sometimescalledIF-THENexpression,itexpressesacondition-resultform,whichisarelativelysimplewayofexpressingknowledge.ThefollowingpartofIFdescribesTheprerequisiteoftherule,andthefollowingpartofTHENdescribestheconclusionoftherule.Theruleexpressionmethodismainlyusedtodescribeknowledgeandstatethecontrolbetweenvariousprocessknowledge,andthemechanismofinteraction.

Example:TheMYCINsystemhasthefollowingproductionknowledge(whereinconfidenceiscalledrulestrength):

IFThestainofthisorganismisGram-negative,theshapeofthisorganismisrod-shaped,andthepatientisanintermediatehost

THENThemicroorganismisPseudomonasaeruginosa,withaconfidenceof0.6

(3)Framerepresentation

FrameisaspecialeventorobjectAcomplexdatastructureinwhichalloftheknowledgeisstoredtogether.Itsmainbodyisfixed,representingafixedconcept,objectorevent,anditslowerlayeriscomposedofslots,whichrepresenttheattributesofeachaspectofthemainbody.FrameworkItisahierarchicaldatastructure.Theslotinthelowerlayeroftheframecanberegardedasasub-frame,andthesub-frameitselfcanbefurtherdividedintosides.Theattributevalues​​oftheslotandthesidearecalledtheslotvalueandthesidevalue,respectively.TheslotvaluecanbeLogicalornumeric,thespecificvaluecanbeaprogram,condition,defaultvalueorasubframe.Interrelatedframesareconnectedtoformaframesystem,orframenetwork.

Example:UseaframeSaidthefollowingearthquakeevents:[VirtualXinhuaNewsAgency,March15th]Yesterday,anearthquakeoccurredinYuxi,Yunnan,causingpropertylossesofabout100,000yuan.Ifthestatisticsdepartmentneedsdetailedlossfigures,pleasecall62332931.AccordingtoexpertsThemagnitudewillnotexceed4,anditisbelievedthatthelocationinnoman’slandwillnotcausecasualties.

(4)Object-orientedrepresentationmethod

Theobject-orientedknowledgerepresentationmethodisAccordingtotheprincipleofobject-orientedprogramming,aformofmixedknowledgerepresentationisformed,thatis,takingtheobjectasthecenter,encapsulatingrelatedknowledgesuchasobjectattributes,dynamicbehavior,domainknowledgeandprocessingmethodsinthestructureofexpressingtheobject.Inthismethod,Thebasicunitofknowledgeisanobject,andeachobjectiscomposedofasetofattributes,relationships,andmethods.Thevalueofanobject’sattributesetandrelationshipsetdescribestheknowledgeoftheobject;themethodsetrelatedtotheobject,Thevalueoftheoperationontheattributesetandtherelationshipsetrepresentstheknowledgeprocessingmethodthattheobjectactsontheknowledge,includingknowledgeacquisitionmethod,reasoningmethod,messagetransmissionmethodandknowledgeupdatemethod.

(5)SemanticWebRepresentation

SemanticWebisoneofthemostimportantmethodsofknowledgerepresentation.Itisahighlyexpressiveandflexibleknowledgerepresentationmethod.ItAnetworkgraphthatexpressesknowledgethroughconceptsandtheirsemanticrelations.Fromthepointofviewofgraphtheory,itisa"markeddirectedgraph".Thesemanticnetworkusesadirectedgraphcomposedofnodesandlabelededgestodescribetherelationshipbetweenevents,concepts,conditions,actions,andobjects.Thelabeleddirectedgraphcandescribetherelationshipbetweenobjectsverynaturally.

Example:Useasemanticnetworktoexpressthefollowingknowledge:XiangyaMedicalCollegeofCentralSouthUniversityisauniversitylocatedinChangshaCity,andwasestablishedin1914.

(6)XML-basednotation

InXML(eXtensibleMarkuplanguage,extensiblemarkuplanguage),dataobjectsaredescribedbyelements,andtheattributesofdataobjectscanbedescribedasTheelement'schildelementsorelementattributes.AnXMLdocumentiscomposedofseveralelements,andtherelationshipbetweendataisembodiedinthenestedformofparentelementsandchildelements.IntheprocessofXML-basedknowledgerepresentation,XMLDTD(DocumentTypedefinitions,documenttypedefinitions)isusedtodefineagrammaticalsystemofknowledgerepresentationmethods.InterprettheinstantiatedknowledgerepresentationdocumentthroughacustomXMLapplication.Intheprocessofknowledgeutilization,thecontentmarkedbyaspecifictagisparsedbymaintainingadatadictionaryandanXMLparsingprogram,andthespecificknowledgecontentisexpressedintheformatof"tag"+"content".Knowledgerepresentationisthekeytoconstructingaknowledgebase.Whethertheknowledgerepresentationmethodissuitableornotisnotonlyrelatedtotheeffectivestorageofknowledgeintheknowledgebase,butalsodirectlyaffectstheefficiencyofthesystem'sknowledgereasoningandtheabilitytoacquirenewknowledge.

(7)Ontologyrepresentation

Ontologyisaformal,shared,clear,andconceptualspecification.Ontologycanexpresssemanticsinanexplicitandformalway,improvetheinteroperabilitybetweenheterogeneoussystems,andpromoteknowledgesharing.Therefore,inrecentyears,ontologyhasbeenwidelyusedinthefieldofknowledgerepresentation.Thepurposeofusingontologytoexpressknowledgeistounifytheconceptsintheapplicationdomain,andconstructanontologyhierarchysystemtoexpressthesemanticrelationshipbetweenconcepts,soastorealizethesharingandreuseofknowledgebetweenhumansandcomputers.Fivebasicmodelingprimitivesarethebasiccomponentsoftheontologyhierarchy.Theseprimitivesare:class,relationship,function,axiomandinstance.UsuallyalsoClasses(class)iswrittenasConcepts.Theontologyisintroducedintotheknowledgemodelingoftheknowledgebase,andthedomainontologyknowledgebaseisestablished.Theknowledgecanbeexpressedwithconceptsandtheinherentrelationshipbetweentheseknowledgescanberevealedatthesametime.Theknowledgeinthedomainontologyknowledgebaseisnotonlyclassifiedthroughtheverticalcategory,butalsoorganizedandrelatedthroughthesemanticassociationoftheontology.Theinferenceenginereusesthisknowledgeforreasoning,therebyimprovingtherecallandaccuracyofretrieval.

Theabovebrieflyintroducesandanalyzescommonknowledgerepresentationmethods.Inaddition,therearesomeknowledgerepresentationmethodssuitableforspecialfields,suchas:conceptmap,Petri,grid-basedknowledgerepresentationmethods,roughsets,Theknowledgerepresentationmethodofcloudtheory,etc.,willnotbeintroducedindetailhere.Intheactualapplicationprocess,anintelligentsystemoftencontainsmultiplerepresentationmethods.

Thecompleteprocess

Inageneralsense,knowledgerepresentationisasetofconventionsfordescribingtheworld,whichisthesymbolization,formalizationormodelingofknowledge;fromcomputerscienceFromthepointofview,knowledgerepresentationisageneralmethodtostudythefeasibilityandeffectivenessofcomputerrepresentationofknowledge.Itisastrategytorepresenthumanknowledgeasadatastructureandsystemcontrolstructurethatcanbeprocessedbyamachine.

Acompleteknowledgerepresentationprocessis:first,thedesignerdesignsavarietyofknowledgerepresentationmethodsforvarioustypesofproblems;thentheuseroftherepresentationmethodselectsasuitablerepresentationmethodtorepresentacertaintypeofknowledge;finallytheknowledgeTheuserusesorlearnstheknowledgeprocessedbytherepresentationmethod.Therefore,theobjectofknowledgerepresentationisknowledge;thesubjectofknowledgerepresentationincludesthreecategories:designersofrepresentationmethods,usersofrepresentationmethods,andusersofknowledge.Specifically,thesubjectofknowledgerepresentationmainlyreferstopeople(individualsorcollectives),andsometimescomputers.

TheprocessofknowledgerepresentationisshowninFigure1."KnowledgeI"inFigure1referstotacitknowledgeorexplicitknowledgeexpressedusingotherrepresentationmethods;"KnowledgeII"referstoexplicitknowledgeexpressedusingthisknowledgerepresentationmethod.Thedeepstructureof"KnowledgeI"and"KnowledgeII"arethesame,buttherepresentationisdifferent.Therefore,theprocessofknowledgerepresentationistheprocessoftransformingtacitknowledgeintoexplicitknowledge,ortheprocessoftransformingknowledgefromoneformofrepresentationintoanother.

Relatedexamples

Supposethereissuchaknowledgethatneedstobeexpressed:XiaoPanisastudentintheDepartmentofComputerScience,buthedoesnotlikeprogramming.Ifweusefirst-orderpredicatelogictoexpressit,weneedtotakethefollowingsteps:First,definethepredicate

Computer(x):xisastudentoftheDepartmentofComputerScience

Like(x,y):xlikesy

Secondly,itisexpressedbyapredicateformula:

Computer(xiaopan)∧¬Like(xiaopan,programing)

Productionrules:Amethodofexpressingknowledgeusedinthejudgmentofconditions,causationandothertypes.ThebasicformisP→Q,orifPthenQ.Herethisproductionruleisdifferentfromtheexpressionof"implication(→)"inthepredicatelogicjustnow.Thelatterisanexactmatch,thatis,ifx,100%willbey,whiletheformercanbeRepresentsafuzzymatchwithacertaindegreeofconfidence,thatis,theprobabilityofoccurrence.

Forexample:if"coughandfever",then"cold",theconfidencelevelis80%.Heretheifpartindicatestheconditionpart,thethenpartindicatestheconclusionpart,andtheconfidencedegreeindicatestheprobabilityofobtainingaconclusionwhentheconditionismet.Thiswholepartformsarulethatexpressesthiskindofcausalknowledge:"Ifthepatienthasafeverandcough,thenheislikelytohaveacold."

Therefore,formorecomplexsituations,wecanusethisknowledgerepresentationofproductionrulestoformaseriesofrules.