ofconcept-representativesymbols Positionpaper Jo˜aoMiguelCunha,PedroMartins,Am´ılcarCardoso,PenousalMachado DepartamentodeEngenhariaInformticadaUniversidadedeCoimbra{jmacunha,pjmm,amilcar,machado}@dei.uc.pt Abstract.ThevisualrepresentationofconceptsorideasthroughtheuseofsimpleshapeshasalwaysbeenexploredinthehistoryofHumanity,anditisbelievedtobetheoriginofwriting.Wefocusputationalgenerationofvisualsymbolstorepresentconcepts.Weaimtodevelopasystemthatusesbackgroundknowledgeabouttheworldtofindconnectionsamongconcepts,withthegoalofgeneratingsymbolsforagivenconcept.Wearealsointerestedinexploringthesystemasanapproachtovisualdissociationandvisualconceptualblending.ThishasagreatpotentialintheareaofGraphicDesignasatooltobothstimulatecreativityandaidinbrainstorminginprojectssuchaslogo,pictogramorsignagedesign. Keywords:Computationalcreativity,Computationalgeneration,Conceptrepresentation,Visualrepresentation 1Introduction Creativitycanbeseenastheabilitytocreatenovelideasbymakingconnectionsbetweenexistingones.ItplaysanimportantroleintheareaofGraphicDesignnotonlyinconceivingnewconceptsbutalsoinvisuallyrepresentingthem. Asfarasvisualrepresentationofconceptsisconcerned,humanshavebeendoingitsincemorethantwohundredthousandyearsago–takeforexamplecavepaintings.Theserepresentationsvaryfrompletelypictorial–e.g.pictograms–tomoreabstract–e.g.ideographs. Thelinkbetweenthevisualrepresentationandtheconceptualconnectionsbehinditcaninfactbeobserved.ExamplesofthiscanbeseenbylookingatChinesecharacters,morespecificallyattheonescategorisedasIdeogrammicCompounds(seeFigure1).Thesecharacterscanbeposedintoothers,whoseconceptsaresemanticallyrelated,belongingtothesame(oratleastsimilar)conceptualspace[6]. Someauthorswereinspiredbythisrelationshipbetweenconceptstotheirvisualrepresentations.OneofthemwasCharlesBlisswhodevelopedmunicationposedofseveralhundredsofideographsthatcanbinedtomakenewones–Blissymbols[1].Inhissystemthevariationintermsofabstractiondegreecanalsobeobserved(seeFigure2). Copyright©2015forthispaperbyitsauthors.Copyingpermittedforprivateandacademicpurposes.InProceedingsoftheICCBR2015Workshops.Frankfurt,Germany. 157 Fig.1.Chinesecharactersforroot,tree,woodsandforest(lefttoright).Rootcanbeobtainedbyaddingalinetothetreecharacter;woodscharactercanbeobtained(barely)byusingtwotreecharacters;woodscanbeobtainedbyusingthreetreecharacters. Fig.2.Blissymbols.Severalinterestingthingscanbeobservedbylookingatblissymbols:suchasavariationintermsofabstractiondegree(therearebothpictorialandabstractsymbols);biningsymbols,newmeaningsareobtained(examplesinFigure2:pen+man=writer,mouth+ear=language);byusingthesamesymbolsinadifferentposition,anewmeaningisobtained(seesymbolswater/rain/steam/stream). Inspiredbyexamplesastheonespresentedabove,ourgoalistoconceiveanapproachputationallygeneratingconcept-representativesymbols,i.e.visualrepresentationsofconcepts.Inthispaperwepresentsomeofthekeyaspectsthathavetobeconsideredwhengeneratingsuchsymbolsandthestrategiestoexploreinordertoachieveourobjective. 2Generationofconcept-representativesymbols Theideaofcreatingasymbolforagivenconceptbasedonitsconnectionstootherconceptsis,justbyitself,interesting.However,ifweconsidertheexplorationofthisideaputationalmeanstoautomatethegenerationprocessofthesymbols,itspotentialgreatlyincreases. Wecanthinkofatoolcapableofgeneratingsymbolswhosevisualpropertieswouldbetheeofananalysisoftheconceptualspaceoftheintroducedconcept.Webelievethatsuchatoolcouldassistthedesignerduringtheideationprocessbystimulatingitscreativity,aidinginbrainstormingactivitiesandthusgivingrisetonewideasandconcepts. Concerningthevisualqualitiesofthegeneratedsymbols,itiscrucialtoconsiderseveralaspects.Thefirstoneisthedegreeofabstraction.Thisaspectcanbeconsideredtobeinfluencedbythechoiceoftheconnectionsusedinthesymbolgeneration.Takeforexampletheconceptcar:ifweconsidertheconnectionsbetweencarandtheconceptsdoor,windowandwheel,theresultantsymbolwillprobablybehighlypictorial;ifwechoosetoignorethoseconnections, 158 theresultingsymbolmightbemoreabstract.Ideally,thetoolshouldallowtheabstractiondegreetobesetordingtotheuser’sneeds. However,thisisnottheonlyaspectthatisgreatlydependentontheconnectionsused.Asobservedintheblissymbols,abinationofsymbolsmightleadtodifferentinterpretations.Ifsomeperceptualaspectsarenotconsidered,thismightresultinaconflictbetweentheconceptandtheperceptionofthesymbol.Inthenextsubsectionsomeoftheseaspectswillbepresented. 2.1Consideringvisualcharacteristics Whendealingwithsymbols,itisimportanttobearinmindsomeaspectsofhowarepresentation’smeaningcanbechangedbychangingsomeofitsvisualcharacteristics.Thefollowingaspectsareessential:position,colourandshape. Thefirstsemioticaspect–position–canbeseeninFigure2.Byputtinganarrownexttothewatersymbolanewconceptisrepresented.Theconceptalsovariesordingtopositioningofthearrow.Suchdetailsmustbeconsideredandamechanismforanalysingthemneedstobedeveloped(similarissueshavebeenconsideredin[2]). Fig.3.Leftsideisshownhowthemeaningofabananacanchangewithitscolour(mature,greenandredbanana).RightsideisKiki/Boubaexample.ordinglytoRamachandran98%ofallrespondentsatributethenameKikitotheshapeontheleftandBoubatotherightone,despitehavingnomeaningatall[9].Bestviewedincolour. Anotheraspecttoberegardediscolour.ThroughabriefanalysisofthebananaexampleinFigure3itiseasytounderstandtheimportanceofthisaspect.Bysimplyattributingadifferentcolourtothesamesymbol,itsperceptualmeaningalsochanges.Inaddition,theuseofcolourhasalreadybeenprovenasameanoffacilitatingtheinterpretationofvisualrepresentationsofconcepts(e.g.[7]).However,itsincorrectusehastheoppositeeffect(e.g.Stroopeffect),causinginterferenceinitsinterpretation.Ontheotherhand,amechanismtoavoidanover-useofcolourwillprobablybeneededascolourmightnotbenecessaryinsomesymbols. Thethirdimportantaspectisshape.Whengeneratingvisualsymbolsfromtextualdata(e.g.works),onecannotavoiddealingwithshape.Thechoiceofshapeforagivenconceptisnoteasybyitselfbutonehasalsotoconsideritsvisualqualities.TakeforexampletheshapespresentedinKiki/boubaexampleinFigure3.Despitenothavinganymeaningatall,thereisacleartendencyorbiaswhenattributingnamestothem.Thiscanbeexplainedasfollows,humans 159 tendtoperformmappingsamongdomains,namelybetweenimageandsound,assuchsharpshapestendtobeassociatedwithsharpsoundsanicshapeswithsmoothones[9]. Aswehavealreadymentioned,thesesemioticaspectshavetobeconsideredwhengeneratingsymbols.Thisisonlypossibletoachievebythoroughlyanalysingtheworkandalsoconsideringpreviouslygeneratedsymbolsasbothexamplesandbaseforthegenerationofthenewsymbol. 2.2Gettinginformation Anextensiveanalysisoftheconceptualspaceisimportantbutthereisanotherissuethathastoberesolved:ifthesystemdoesnothaveesstoalargesourceofknowledge–withinformationaboutvisualproperties–itwillbedifficult,ifnotimpossible,toachievegoodresults.Onepossibilityistouseanalreadybuiltworkmonsenseknowledge(e.g.[8]or[3]). However,asourmainobjectiveistogeneratevisualrepresentations,knowledgeaboutvisualcharacteristicsisrequired.Forthatreason,amethodologyhastobeconceivedforacquiringsuchdata.ApossiblesolutionistouseasimilarapproachtotheoneusedbyOpenMindCommonSenseproject–aknowledgeacquisitionsystemdesignedtomonsenseknowledgefromthegeneralpublicovertheweb[10].Ourgoalistofocusongatheringinformationaboutobjects’visualcharacteristicssuchascolour,shapeandtexture.Thesewilllikelyallowustoattainadequateresultsintermsofsymbolgeneration. Crowdsourcingwillprobablybeusedinourknowledge-gatheringprocessasiteasilyallowstoreachahighnumberofcontributorsatareducedcost.Inaddition,thevalidityofonlinecrowdsourcedexperimentsonvisualpropertiesandgraphicalperceptionhasalreadybeendemonstrated(e.g.[5]). Thisdistributedhumanprojectapproachallowsusnotonlytogatherdataatascalethatwouldnotbepossibleotherwisebutalsoenablesustostudytheroleofcontextinperception–oneofourgoalsistotestwhetherthesymbolsgenerateddifferordinglytothelocationwherethedatawasgatheredfrom. Wealsointendtoexploreotheralternativesforpopulatingourwork,suchasautomaticgatheringofinformation.UsingGoogleImageSearchisoneexampleofthisandcanbeusedtofindimagesrelatedtothecontentbeinganalysedandconsequentlyextractingusefulinformationfromthem(e.g.[7]). 2.3Generatingsymbols Inouropinionthereare,atleast,twodifferentwaysofgeneratingasymbolforagivenconcept:(1)startingwithnopriorknowledgeandanalysingtheconceptualspaceinordertoextractpossiblevisualfeaturestobeusedinthesymbol;(2)usingpriorknowledge–previouslygeneratedsymbolsofconceptsthatare,insomeway,relatedtotheintroducedconcept.Thiswouldleadtoahighercoherenceamonggeneratedsymbols.Inbothcases,notonlydirectconceptualconnectionsareusedbutalsomoremonones–throughaprocessofanalogy.Assuch, 160 wearguethatmechanismssuchascase-basedreasoningorconceptualblending[4]aresuitablestrategiestogeneratesymbolsofconcepts.Asfortheformer,wecanconsideraprisedofsymbolssuchastheonesdepictedinFigure2anddevelopasystemtoproducenovelonesusinganaloguestothepreviousones.Regardingconceptualblending,theideaistoexplorethestructuremappingapproachtoanalogyandconceptintegrationbasedonconceptualspacesandsemioticsystems. 3Conclusion Inthispaperwepresentedourapproachputationalgenerationofconceptrepresentativesymbols.Weaimtodevelopadesignaidingtoolbinestheexplorationofconceptualspacesbinationwithprocessesofanalogymakingandsemioticanalysistogeneratepossiblevisual(abstract/semi-abstract)representationsfortheconceptsintroducedbytheuser.Webelievethatitwillhelpthedesignerduringtheideationprocessbystimulatingitscreativity,aidinginbrainstormingactivitiesandthusgivingrisetonewideasandconcepts. AcknowledgementsTheauthorsacknowledgethefinancialsupportfromtheFutureandEmergingTech- nologies(FET)programmewithintheSeventhFrameworkProgrammeforResearchoftheEuropeanCommission,undertheConCreTeFET-Openproject(grantnumber611733)andthePROSECCOFET-Proactiveproject(grantnumber600653). References
1.Bliss,
C.K.:Semantography(Blissymbolics):ALogicalWritingforanillogicalWorld.SemantographyBlissymbolicsPubl(1965)
2.Confalonieri,
R.,etal.:Usingargumentationtoevaluateconceptblendsbinatorialcreativitypapertype:Studypaper.(toAppearinICCC2015)
3.DeSmedt,
T.:ModelingCreativity:CaseStudiesinPython.UniversityPressAntwerp(2013)
4.Fauconnier,
G.,Turner,
M.:Conceptualworks.Cognitivescience22
(2),133–187(1998)
5.Heer,
J.,Bostock,
M.:Crowdsourcinggraphicalperception:usingmechanicalturktoassessvisualizationdesign.In:ProceedingsoftheSIGCHIConferenceonHumanFactorsinComputingSystems.pp.203–212.ACM(2010)
6.Hew,
S.,etal.:biningevolutionofpictogramchinesecharacterstorepresentpoundschinesecharacters.In:ComputingandConvergenceTechnology,20127thInternationalConferenceon.pp.219–223.IEEE(2012)
7.Lin,
S.,etal.:Selectingsemantically-resonantcolorsfordatavisualization.In:ComputerGraphicsForum.vol.32,pp.401–410.WileyOnlineLibrary(2013)
8.Liu,
H.,Singh,
P.:amonsensereasoningtool-kit.BTtechnologyjournal22
(4),211–226(2004)
9.Ramachandran,
V.S.,Hubbard,
E.M.:Synaesthesia–awindowintoperception,thoughtandlanguage.Journalofconsciousnessstudies8(12),3–34(2001) 10.Singh,
P.,etal.:Openmonsense:Knowledgeacquisitionfromthegeneralpublic.In:Onthemovetomeaningfulsystems2002:Coopis,doa,andodbase,pp.1223–1237.Springer(2002)