chapter1 计算机1

1PART1METHODOLOGIES&SKILLSChapter1HowtoMakeaResearchPresentation2ThinkAbout…Whendoyouneedtomakearesearchpresentation?Haveyoueverhadanexperienceofmakingapresentation?Howdoyoufeelaboutit?Whatisthecommonstructureofapresentation?Andwhichpartsareimportantandmustbeexplainedclearly?Whatshouldyoupayattentiontowhenyougiveapresentation?Isthereanythingelsethatyouthinkisalsoimportantbuthasnotbeenmentionedinthetextbookforapresentation?3WhenWeNeedtoMakeaPresentation•Usuallyweneedtopresentourresearchinaregularseminarorataconference.4WhatistheClearStructureofaPresentation•Titleofthepresentation/name/affiliationoftheauthor•Outline•Why/Whatyouwanttodo•Priorart/relatedwork/what’snew•Howtodo•Results•Evaluation:contributions/advantages/limitations•Applications/recommendations•Conclusion/futurework5Sample:SocialAvatarLanguageforNaturalAvatar-BasedHumanCommunicationDr.JinHouSouthwestJiaotongUniversity6Outline•WhatandWhywewanttodo•Relatedworkandchallenges•Ourapproach•Result•Conclusionandfuturework7SocialAvatarLanguage(SAL)•Avatar?–Graphicalrepresentationofarealpersoninavirtualenvironment•SAL?–Avatars’nonverballanguageintermsofsocialcontext•WhySAL?–Ourbehaviorsareinfluencedbysocialfactorssuchasculture,profession,religion…•Purpose!–Natural,sociable,personalizedavatars8Challenges•Whereistheavatarfrom?•Whatisthepersonalityoftheavatar?•Bodymovementismostlyneglectedandnotsynchronizedwiththeface•Commercialproductsareavailablebutusuallyrelyonthepre-authoredaudio9WhatHasBeenDonewithAvatars•MITMediaLab,JustineCassellgroup–BEAT(BehaviorExpressionAnimationToolkit)•PennsylvaniaUniv.,NormanI.Badlergroup–EMOTE(ExpressiveMOTionEngine)•Other:USC,NYU,F2F,Sony…10BEAT•Advantages:–Currentlythemostsatisfyinganimationtoolkitthatautomaticallyextractsactuallinguisticandcontextualinformationfromtexttosuggestappropriatehandandarmgestures,facialexpressions,andintonationofvoice–Allowsanimatorstoinserttheirpreferredpersonalities,motioncharacteristics,orotherparticularfeaturesinthefinalanimation•Disadvantages:–TheimperfectcomputationallinguisticsisthegreatestlimitationinBEAT’scurrentversion–OnlyperformsforEnglishinput11EMOTE•Advantages:–Offersamethodologicaltoolforcreatingpersonalizedagents,whichprovidesamethodofsynthesizingmovementsbycontrollingarelativelysmallnumberofparameters,andenablesahigherlevelofunderstandingcommunicativegestures–Acquiregestureswithemphasisonarmandbodyposturesthatarelargelyneglectedbyothersystems•Disadvantages:–Notprovideaneffectiveandeasy-to-usetoolforaddressingthesocialcontextforavatars–Notconsidermultilingualinput12What’sNewinOurStudy•Newconception:SALtoreflectthesocialbackgroundofanavatar•Emphasisonculturalfactor:Japanese,Chinese,American•Includingbodymovement,facialexpressionswithappropriateeyes,eyebrows,andmouthmovements13OurApproach-Ontology+VRML•VRML-basedH-AnimHumanoid•OntologyModel–AvatarModelingOntology(AMO):putstheanatomicconstraintsonavataranimation–SocialContextOntology(SCO):associatesthesocialcontextwithavatars14AvatarModelingOntology(AMO)•Definemeta-elements(segment,joint…),andfurtherelementslikebodyparts(arm,leg…)andfacialexpressions(happiness,fear,disgust…)–Identity:distinguishesanentityfromtheothersintermsofgeometry,appearance,texture,colorandsoon–Dependency:describesalloftherelationshipsanentityhas,e.g.theupperarmadjoinstheforearm15SocialContextOntology(SCO)•Associatethesocialcontextwithanordinaryavatar-cultureontology,professionontology,familyontology,religionontology–Definition:asetofontologyelements{Ψi|i=1,…,n},whereΨi=vi,R’i,Ii,Idi–RelationshipGeneralization-Φgeneralization(Asian,Japanese)Disjoint-Φdisjoint(teacher,student)–ExtensionMechanism:change/adapttheexistingontologyorintroduceanewontology16Client-ServerTopologyQuerySCORetrievalDatabaseServerClientCommonGatewayInterface(CGI)MappingManagerHTMLFileVRMLFileSemanticFeatureExtractionSCOProfileUserInterfaceVRMLBrowserCharacterDBTransformDBAnimationDBSCODBKeywordDB17FlowChartofAnimationSynthesisInputOutputNaturalLanguageProcessingAvatarWordKeywordDBTransformDBAnimationDBCharacterDBVRMLAnimationSynthesisHTMLGenerationSCODB18私(は)正座(しています)(Iamsittingonmyheels)19我去游泳(Igoswimming)20IGoSwimming21FacialAnimation22Application（1）•Graphicalchat23GraphicalChat•Displayconceptsthatwouldbeverydifficulttoexplainverballytopeoplefromdifferentcultures,suchas“seiza”inJapan,orTaiChiinChina•Atoolforenhancinginternationalunderstandingacrossculturalbarriersbyusingbothverbalandnonverbalchannels•Currentspokenlanguagemachinetranslationsarenotsatisfying,ourSALcansupplementmachinetranslationandimproveunderstandingthroughprovidingavisualaid24Application（2）•Virtualinterpreter25VirtualInterpreter•Increasebelievabilityandrealitybybothverbalandnonverbalbehaviorinvolved,especiallywithemotionalexpressions•Applicationtomanyfieldslikeweatherreporter,businessorotherserviceassistant,pedagogicalagent…26Conclusion•Weoriginallyproposetouseontologyforaddressingthesocialcontextforavatars•Integratesfacialexpressionswithbodygestures•Effectivemulti-naturallanguageprocessingalgorithm•Thissystemisindependen