Background
Intheresearchofartificialintelligence,commonsenseknowledgeisthefactscollectedandtheinformationthatanordinarypersonpredicts.Theproblemofcommonsenseknowledgeisanongoingprojectthatbelongstothefieldofknowledgeexpression(abranchofartificialintelligence).Itaimstocreateaknowledgebaseofcommonsense:adatabasecontainingallthegeneralknowledgethatmostpeoplehave.Itisexpressedinsuchawaythatitcanusenaturallanguageinartificialintelligenceprogramsormakeinferencesabouttheordinaryworld.Suchadatabaseisakindofontology,themostcommonofwhichistheso-calledupper-levelontology.
Thisquestionisconsideredthemostdifficult,becausethebreadthofartificialintelligenceresearchanddetailedcommonsenseknowledgeishuge.Anytaskthatrequirescommonsenseknowledgeisconsideredtobeaccomplishedbyartificialintelligence:alotofworktobedone,andwhatapersondoesnothave,itrequiresamachinetoshowtheintelligenceofaperson.Thesetasksincludemachinetranslation,objectrecognition,textmining,andmanyotherproblems.Toperformthesetasksperfectly,themachinesimplyknowswhatthetextistalkingaboutandtheobjectisvisible.Thisisimpossibleunlessthemachineisfamiliarwiththesameconceptsasanordinaryperson.
Someexpertsbelievethat"common-senseknowledgeprocessingisthecoreproblemofartificialknowledgeresearch."Someexpertsbelievethat"howtoeffectivelyacquiretheknowledgeofexpertsinthefieldhasalwaysbeenaprobleminartificialintelligence."
Definition
Thedomainknowledgebasepossessedbyearlyartificialintelligenceorknowledgeengineeringsystemsisanotherkindofknowledgebase.Inotherwords,domainknowledgebaseandcommonsenseknowledgebasearethetwobasictypesofknowledgebasepossessedbyintelligentcomputersystem.Thefieldofcomputersciencegenerallybelievesthatdomainknowledgebasesandcommonsenseknowledgebasesarethebottleneckproblemsofartificialintelligenceorknowledgeengineeringtechnology.Fromfocusingonthedomainknowledgeofexpertsintheearlydaystonowfocusingoncommonsenseknowledgeatthesametime,thisisanadvancementinartificialintelligenceorknowledgeengineeringtechnology.Duetothecontinuousmaturityofcomputerhardwareandsoftware,databasesandevendatawarehousesandtheirhuman-computerinteractioninterfacetechnologies,peopleneedtodevelopvariousmedium-scaledomainknowledgebasesandlarge-scalecommonsensesystemsrequiredforthedevelopmentofvariousexpertsystemsinthe21stcentury.Alarge-scalecommonsenseknowledgebasehasthebasicconditions.
Listofcommonsensecontent
Theinformationinthecommonsenseknowledgebasemayincludebutisnotlimitedtothefollowing:
ClassandontologyIndividual
Materialofpartsandobjects
Objectattributes(suchascolorandsize)
Thefunctionandpurposeoftheobject
Thelocationanddistributionoftheobject
Thelocationofactionsandevents
Timeofactionsandevents
Prerequisitesforactionsandevents
Theimpactofactionsandevents(post-conditions)
Thesubjectandobjectoftheaction
Thebehaviorofthedevice
Theobsolescenceofthesituationorscript
Humanrightsgoalsandneeds
Emotion
Planandstrategy
Storytheme
Context
PopularExamples
WordNet,whichisadictionary-styleknowledgebaseofcommonsense.
(Large-scale)CommonSenseKnowledgeBase(Cyc),whichisalarge-scalecommonsenseknowledgebasesimilartoanencyclopedia.
MindCollection,itisaknowledgebaseofcommonsenseinvolvingnaturallanguageprocessing.
TheSemanticWeb,whichisafuturenetworkwithbothadomainknowledgebaseandacommonsenseknowledgebase.
OpenMindCommonSense
Basicformalontology
Generalformalizedontology
Conceptnetwork
Thinkingdotmap
Ontology
Upperorupper-levelontology
Commonsensereasoning
Developmentdirection
Practicallyresearchtheontologystructureandoperatingmechanismofalarge-scalecommonsenseknowledgebaseforintelligentagents.Intheinternationalartificialintelligencecommunity,ithasalwaysbeenrecognizedthattheprocessingofcommonsenseknowledgeisthecoreproblemofartificialintelligence.Theso-calledcommonsenseisrelativetoprofessionalknowledge.Professionalknowledgeiswidelyusedinvariousexpertsystemsandapplicationsoftware.Althoughtheprofessionalknowledgeaccumulatedbyhumanbeingsisvast,butcomparedwithcommonsense,professionalknowledgeisonlyalittlewitch’sopinion.Greatwitch.McCarthyoncepointedout:Lackofcommonsenseisamajorweaknessofthecurrentexpertsystem.Averyimportantresearchdirectionofcommonsenseproblemsisthemathematicalmodelofcommonsenseandcommonsenseprocessing,whichinvolvesvariousrepresentationsofcommonsense(first-orderlogic,productionsystems,semanticnetworks,neuralnetworks,etc.),aswellasvariouscommonsenseprocessingPrinciples(deduction,induction,qualitativereasoning,fuzzyreasoning,non-monotonicreasoning,sub-consistentreasoning,etc.),especiallynon-monotonicreasoningandnon-monotoniclogic,arespecificallyproposedforthestudyofnon-monotonicityincommonsensereasoning.AboutcommonsenseproblemsVarioustheoreticalstudieshavegreatlypromotedtheprogressofcommonsensereasoning,someofwhichhavebeguntodevelopinpracticalaspects,suchastheuseofnon-monotonicreasoningfortruthmaintenance.However,thesestudiesarestilltheoreticalstudiesintermsoftheirmainaspects.Thelackof"heavyweight"applicationsismainlyduetothewide,vagueandcomplexcharacteristicsofcommonsenseanditsreasoning.Forexample,Sowacomparescommonsenseto"knowledgesoup".CaoCungenlikenscommonsensetoacloudofknowledge.Thestudyofnon-monotoniclogiconlycapturesthefeatureofnon-monotonicityincommon-sensereasoning.Infact,therearemanyvariouscharacteristicsincommon-sensereasoningthatneedtobestudiedextensively.
