writtengooglevoice

It'snotWhatItSpeaks,butIt'sHowItSpeaks:AStudyintoSmartphoneVoice-UserInterfaces(VUI)JaeyeolJeongandDong-HeeShin(&)DepartmentofInteractionScience,SungkyunkwanUniversity,Seoul,Korea{jae10,dshin}@skku.
eduAbstract.
Sincevoice-userinterfaces(VUI)arebecominganattractivetoolformoreintuitiveuserinteractions,thisstudyproposesabetween-subjectexperi-mentinwhichvariationsinvoicecharacteristics(i.
e.
,voicegenderandmanner)ofVUIareexaminedaskeydeterminantsofuserperceptions.
Thisstudypre-dictsthatthevoicegender(malevs.
female)andmanner(calmvs.
exuberant)arelikelytohavesignicanteffectsonpsychologicalandbehavioroutcomes,includingcredibilityandtrustworthinessofinformationdeliveredviaVUI.
Keywords:VoiceuserinterfaceVoicegenderVoicemannerSmartdeviceCredibilityTrust1IntroductionAssmartdevices(e.
g.
,smartphones,smartwatches,smartTVs)havebecomeubiqui-tous,anincreasingemphasisisbeingplacedonvoice-userinterfaces(VUI),suchasAppleSiriandGoogleVoiceSearch,asaneffectivesolutiontoprovidinggreaterinteractivityandmorepositiveuserexperience.
VUIenablesuserstointeractwithcomputersthroughthevoicerecognitionandcommand.
Itoffersmoreconvenient,safe,anduser-friendlyinterfacesthantheconventionaltouch-basedinterface,allowingseeminglyhands-andeyes-freeinteractionsviathedevice.
Voiceisanintegralcomponentofhuman-humaninteractionbecauseitreectsandconveysemotions,intentions,andmannersofthespeaker.
Judgmentsareoftenmadesimplybasedonhowcredibleandbelievablethespeaker'svoicesounds[1].
But,doesthevoicealsomatterinhuman-computerinteractionsDoesitinuencehowusersevaluatethecredibilityoftheinformationdeliveredbydigitalmediaanditssourceTheComputersAreSocialActors(CASA)paradigmsuggeststhathuman-humanandhuman-computerinteractionssharesimilarcharacteristics,suchthatusersmindlesslyapplythesamesocialrulesandexpectationswhenusingcomputerswithsocialcuesastheywouldinhuman-humaninteractions[2].
Giventhatvoiceisahumanlikeattributethatfunctionsasastrongsocialcue,theCASAparadigmisalsolikelytobeapplicabletoinvestigatingtheroleofvoicecharacteristicsininducinggreatercredibilityofinformationdeliveredviaVUI.
Ampleresearchonhuman-computerinteractionhasdemonstratedthatvariationsinvoicecharacteristicsleadtodifferentuserattitudeandperception.
Forexample,SpringerInternationalPublishingSwitzerland2015M.
Kurosu(Ed.
):Human-ComputerInteraction,PartII,HCII2015,LNCS9170,pp.
284–291,2015.
DOI:10.
1007/978-3-319-20916-6_27individualsarefoundtoapplygenderstereotypestocomputersandmachines.
Productdescriptionsprovidedbymalevoiceareperceivedasmorecrediblethanthosewithfemalevoice[3].
Womenwithmasculinevoiceareperceivedtobemorerationalandpersuasivethanthosewithfemininevoice[4].
Inaddition,masculinevoiceisratedasmorecompetentthanfemininevoice,regardlessoftheactualgender[5].
Giventhatsuchstereotypicalvocalcuesarefoundtoaffectthepersuasiveness,credibility,andcompetenceofthedeliveredmessages[6],theyarealsolikelytohavenotableeffectsonthewaysinwhichusersevaluateandperceiveinformationconveyedviaVUI.
Therefore,thisstudyintendstoexplorewhethervariationsinvoicegender(malevs.
female)andvoicemanner(dynamicvs.
calm)ofVUIcontributetoaffectingpsychologicalandbehavioraloutcomes.
2LiteratureReviewandRelativeWorksVoiceInterface.
Popularityofvoiceinterfacesonsmartphoneshavebeengrowingrecently.
Asformostuserstypingonaphoneismorelaboriousthantypingonafull-sizekeyboard,voiceinterfacesservearealpracticalpurpose.
Voicerecognitiontechnologyisnothingnew–havingbeenaroundsincethe1950swhenintroducedbyBellLabs.
However,itwasn'tuntilthe1990swhenIVRsystems(interactivevoiceresponse)systemsbecamemorewidespreadthatthetechnologyreallybegantosolidify.
Acertainnumberofstudiesaboutaudiotechnologyinsmartdevicehavebeenconducted.
Thisincludesspeechrecognition[7],soundrecognition[7,8],speechsynthesis[9]ordialogue[10–12].
VoiceinterfaceshavegainedpublicrecognitionsincethereleaseofSirioniPhone.
Siri'spopularityisanexcitingdevelopmentforspeechtechnologies.
ManyvoiceinterfaceswithfunctionalitiessimilartoSiri'sarealsoavailableonAndroidsmart-phones,includingGoogle'sVoiceSearch,SpeakToItassistant,VlingoAssistant,Jeannie,andEva.
GoogleAndroidSearchinterfacesthatfocusexclusivelyonsearchfunctionsincludeDragonandBingsearch.
AnexampleofApple'sSiri,whichcanprocessauser'sspeechinnaturallanguage,replytheuserwithinareasonableperiodoftimeandperformroutinetasks.
iPhoneuserscanmakereservationsatspecicres-taurants,buymovieticketsorcallataxibydictatinginstructionsinnaturallanguagetoSiri.
Userscanalsoposesimplequeriessuchas"Whatistheweathertomorrow"WhyVoicemakesSenseasaUserInterface.
Asconsumersgrowmorecomfortablewithvoicerecognitiontechnologyforeverydaytasks,developersareputtingspeechintoavarietyofdevicesincludingremotecontrols,carsandwearabletechnology.
Voiceasauserinterfacemakessensebecauseitissomethingthatcomesnaturallytohumansandisusedonadailybasis.
AccordingtoresearchconductedbyMatthiasMehl,theaveragenumberofwordsspokenbyanadultperdayis16,000,whiletheaveragenumberofwordstypedperdayis3,000to4,000[13].
Akeyadvantageisthefactthatvoiceisnotonlyhands-free,buteyes-free,makingitsuitableforuseinavarietyofenvironments,aswellasforthevisually-disabled.
Andunlikemostwrittenwords,voicecancommunicatemood,gender,identity(suchasrecognizingavoice),emphasisandevenpersonality.
It'snotWhatItSpeaks,butIt'sHowItSpeaks285VoiceInterfaceTypes.
Voiceinterfaceshelpusersperformfunctionssuchassendinganemailoratextmessage,playingasongfromamusiclibrary,accessingcalendar,performingawebsearch,orcheckingweatherforecast.
Theinterfacesdifferinthefunctionsthattheysupport,modalityofcommunication,text-to-speechandrecognitioncomponents,amountofinitiativetakenbythesystem,anddialoguehandlingmethods.
Virtualassistantvoiceinterfacestakearoleofapersonalassistant.
Siri,SpeakToItassistant,andEva/Evanappsspeakbacktoauser.
Theyaddresstheuserbynameandexhibitemotionsbothinachoiceofaspokenresponseandinafacialexpression.
SpeakToItassistantshowshappinesswhenitisturnedonbysaying"Goodtoseeyouagain",apologizeswhentheydonotunderstandorisunabletohandleacommand"IamsorryI'mnotabletodothatjustyetbutIwillbesoon".
WheniPhonelosesnetworkconnection,Siri(whichreliesontheconnection)respondswitherrormessages,suchas"Thereissomethingwrong","Icannotansweryounow".
Theseresponsessoundverycuteandhuman-likeandgiveanimpressionthattheapphasapersonality,howevermorehelpfulresponses"Networkisdown,tryagainlater"or"Tryturningonyourwirelessconnection",wouldrevealtheproblemtoauserorsuggestasolution.
QualityofTTSmakesabigdifferencefortheperceivedqualityofavoiceapp.
NeitherofthemechanicalvoicesusedbythefreeversionsofAndroidsystemscomparestoaffectiveSiri'svoice.
SpeakToItandEvaassistantsalsohaveagraphicalpersona.
AcharacterinSpeakToItassistantapphasacustomizableappearanceanddisplayssubtlefacialexpressionsduringcommunication.
However,itisnotclearhowmuchthisaddstothesystem'sfunctionality.
VoiceSearchInterfaces.
Othertypesofapps,suchasDragonandBingspecializeonsearchonly.
Theseinterfacesarenotattemptingtobe'can-do-all'assistants.
Instead,theyhaveafocusedsetoffunctionsrelevanttosearch.
Binghasapre-setlistofsearchtypes:images,videos,maps,local,etc.
Wefoundthishelpfulbecauseitsuggeststoauserwhichfunctionalitiesaresupportedbythesystem.
Dragon,ontheotherhand,providesaspeaking-onlyinterfaceequivalentto'howmayIhelpyou'.
Ausercanguessbytrialanderrorwhatthesystemcapabilitiesare.
UserEvaluationofSpeechOutput.
Userevaluationofspeechoutputcomprisesthreefactors;theusers'personalpreferences;theoperationalcontext;andoverallsystemfunctionality.
Thesefactorsarenotindependentofeachotherandwillintegratetoinformtheuserintheirevaluationofspeechoutput.
Personalpreferencesforspeechareconstructedthroughtheapplicationofthesocialrepresentationsheldbythelistener.
Socialrepresentationsarecreatedby,forexample,normsandstereotypes,andareemployedtofacilitatecommunicationwithinsocialgroups.
Simplybasedonthe'sound'ofspeech,socialrepresentationsareformedinrelationto,forexample,geo-graphicalorigin,genderandage[14].
Alargeamountofresearchhasfocusedonthesocialrepresentationsgeneratedbyspeechas'human'output,butfewstudiesconsiderthosegeneratedbyspeechassystemoutput.
Thelimitednumberofstudiesthathaveinvestigatedsocialreactionstosyntheticspeechhavediscoveredthatsyntheticspeech,isoftenreactedtoinsimilarwaystonatural(human)speech[15].
Forexample,personalityandgenderareawardedtosyntheticspeechoutputsandgenderstereotypesmayarisesimilarlyforbothnaturalandsyntheticspeech[4–14].
Giventhatspeechoutputisevaluatedonthebasisofuser286J.
JeongandD.
-H.
Shinpreferencesinformedbysocialrepresentations,itappearsappropriatetoconsiderthesmartphoneusers'socialrepresentationsofspeechoutputstodeterminethespokencharacteristicsthatprovokebothpositiveandnegativeuserevaluations.
SpeechGender.
Evenwhenaspeakerisnotpresent,speechisextremelyusefulinidentifyingthegenderofaspeakerandprovokesassociatedsocialrepresentations[15].
Additionally,ReevesandNass[4]arguethateveryoneassignsgendertobothnaturalandsyntheticspeechoutputs.
Theyfoundthatevenwhenthecontentofthespeechoutputwasidentical,malespeechoutputsthatpraisedtheuserwerebetterreceivedthanfemalespeechoutputs;malespeechoutputwasevaluatedasmorefriendlythanfemalespeechoutputs;andmalespeechoutputswereevaluatedasbetterinformationprovidersoncomputers.
Here,itwouldbeinterestingtoinvestigatethecurrentsocialrepresentationsofgenderthatareinvokedinrelationtospeechassmartphoneoutputandtheimpactthatgendermayhaveonsmartphoneuserevaluations.
3ResearchQuestionInhuman-humaninteraction,genderbiasaffectscredibilityofagent.
Forexample,malenewscastersareratedasmorecredibleandcompetentthanfemalecasters[18].
Researchinhuman-computerinteractionhasalsodemonstratedthatusersapplysimilargenderstereotypeswheninteractingwithcomputers[19].
Byextension,thisgenderbiasisalsolikelytobediscoveredwhenexposedtoVUIwitheithermaleorfemalevoicesandaffectuserexperiencewithVUIandperceptionofinformationconveyedviaVUIiftheCASAparadigmisindeedapplicabletothesmartphoneVUIcontext.
Inadditiontovoicegender,voicemannerisanothercriticalvocalcharacteristicthatcontributetoshapinguserperceptions.
Voicemannerislargelydeterminedbyvaria-tionsinpitch,suchthatvoicewithcalmmannerisdenedashavinglowpitchandmonotonesoundswhilevoicewithdynamicmannerisdenedashavinghighpitchandwiderangeofvariedtones.
Researchershavedemonstratedthatvoicewithdynamicmannerisgenerallyperceivedtobemoreenjoyableanduseful[20].
Together,voicegenderandmannerarelikelytoserveascriticalcomponentsofVUIandinuenceuserexperienceandperception.
Thefollowingresearchquestionisaimedatexaminingthispossibility.
RQ:Doesvoicegender(malevs.
female)andvoicemanner(dynamicvs.
calm)ofVUIaffectcredibilityandtrustinmessagesdeliveredviaVUIandenjoymentofinteractingwithVUI4MethodologyandAnalysisA2(voicegender:malevs.
female)*2(voicemanner:dynamicvs.
calm)factorialdesignexperiment(Fig.
1)willbeconducted.
ParticipantswillberecruitedfromalargeprivateuniversityinSeoul,Korea,andpaidvedollarsfortheirparticipation.
It'snotWhatItSpeaks,butIt'sHowItSpeaks2874.
1StimulusMaterialTheresearcherwillselect10restaurantsneartheexperimentsiteandcreateverbaldescriptionsofeachrestaurant.
Inordertoselectrestaurantsthatdonotelicitstrongpositiveornegativereactions,15respondentswillparticipateinapretestthatisdesignedtoratetheirfamiliarityandinvolvementineachrestaurant(e.
g.
,"thisplaceisrelevant/irrelevant,""important/unimportant,"andmeansnothing/meansalottome")[9].
Basedontheresultofthispretest,verestaurantsratedasthemostneutralwillbeselected.
Next,professionalvoiceactorswillbehiredtorecordthedescriptionsoftheselectedrestaurantsinthefourdifferenttypes(i.
e.
,voicegenderxvoicemanner)ofVUI.
4.
2Procedure80undergraduateandgraduatestudentswillberecruitedfromalargeuniversityinSeoul,Korea.
Uponarrivalatthelaboratory,participantswillberandomlyassignedtooneofthefourconditionsandgivenabriefinstructiononoperatingtheVUI.
Par-ticipantswillthenlaunchtheVUIappandaskittorecommendarestaurantneartheuniversity.
TheVUIappwillrecommendoneoftheveselectedrestaurants.
AfterreceivingtherecommendationfromtheVUIapp,participantswillcompleteaques-tionnairewithapaperandpencilmeasuringthefollowingvariables.
4.
3DependentVariablesParticipantswillrespondtothequestionnairebymarkingon7-pointLikertscales(1="stronglydisagree/notatall,"7="stronglyagree/verymuchso").
Credibilitywillbemeasuredwith10itemsadoptedfrom[19]:believable,trustworthy,convincing,credible,reasonable,unquestionable,conclusive,authentic,honest,andlikable.
Trustwillbeassessedbytwoitemsadoptedfrom[20]:"Ihavetobecautiousaboutrec-ommendationmadebythesmartphonebecauseitissomewhatquestionable(reversed)"and"Thesmartphonevoicewaswarmandcaring.
"Enjoymentwillbemeasuredwithtwoitemsadoptedfrom[19]:"UsingthesmartphonewithVUIwasexciting"and"IenjoyedusingthesmartphonewithVUI"[20].
Fig.
1.
Researchmodel288J.
JeongandD.
-H.
ShinParticipantswillalsorespondtothemanipulationcheckitemsthataredesignedtoconrmthattheindependentvariablesaresuccessfullymanipulated:"Thesmartphonehadmale/femalevoice"and"Thesmartphonehaddynamic/calmvoice.
"5DiscussionsandConclusionWhatmakesauserperceivesystemasintelligentSubjectivecharacteristicsplayanimportantroleinuserperception:suchaswittyandvariedresponses,personalizedaddressingtoauser,qualityofTTSandaffectiveintonation.
Objectivecharacteristicsthataffectauser'sperception,besidesqualityofspeechrecognition,includetheabilityofthesystemtocommunicatetoauserwhatthesystemcapabilitiesare,effectiveerrordetectionandhandling.
Voiceinterfacesthattakeinitiativecanbeperceivedasmoreintelligent,howevertakinginitiativealsomeanstakingariskofannoyingauserwithanunnecessaryquestionorrequest.
Thenewpopularityofvoiceinterfacesonsmart-phonesisanexcitingopportunitythatcandriveadvancesindialogueresearch.
Theadvantagesofthevoice-basedinteractiondesignhavebeenhighlightedasthefollowing:–Itprovidesasimple,usableandinterestinguserinterfaceandsatisestheneedformorefreedominahumancomputerinteractionenvironment.
–Itprovidespeoplenewexperienceandgreatpleasurewhichtraditionalinteractioncouldnotoffer.
–Itmakestheinteractionbetweenhumanandcomputermorenatural.
Ithasbeenillustratedinsciencectionmoviesthatthistechnologycanimprovepeople'slivesifitisappliedrightly.
–Itiswidelyusedinvariousapplicationareassinceitgivestheuseranewexperienceoffeeling.
InthispaperwehavestudythetrendsinVoiceUserInterfaceandexperimentaldesign.
Smartdevicesbecomemoreautonomoustheyalsohavebecomeincreasinglypresentinhumanworld.
Moreresearchcanfollowtheupcomingtrends.
Wealsoconcludedthatthebiggestchallengewillbetokeeptheuserengagedandcontinuallyusingthedevice.
Voiceisthemostnaturalandeasiestttokeepconsumersengagedandwillbecomeanecessityasmostwearabledeviceswillbeinincreasinglysmallerformfactors.
Forexample,anumberofsmartphonessuchasMotoXarewidelyincorporatingthistechnologyviaGoogleNowtodelivera"touchless"experience.
Alotoftheassistantshaveveryroboticvoices,butafew(e.
g.
Evi,assistant)haveverynice/realvoices.
Afewhavetheirownvoices(e.
g.
VoiceBrief),andSVOXhascustomvoicesforandroid.
Asspeechsynthesisandspeechrecognitiontechnologiesimprove,applicationsrequiringmorecomplexandmorenaturalhuman-computerinteractionsarebecomingfeasible.
Awell-designeduserinterfaceiscriticaltothesuccessoftheseapplications.
Acarefullycrafteduserinterfacecanovercomemanyofthelimitationsofcurrenttechnologytoproduceasuccessfuloutcomefromtheuser'spointofview,evenwhenthetechnologyworksimperfectly.
Withfurthertechnologicalimprovements,thepri-maryroleoftheuserinterfacewillgraduallyshiftfromafocusonadaptingtheuser'sIt'snotWhatItSpeaks,butIt'sHowItSpeaks289inputtotthelimitationsofthetechnologytofacilitatinginteractivedialoguebetweenhumanandcomputerbyrecognizingandprovidingappropriateconversationalcues.
Amongthefactorsthatmustbeconsideredindesigningvoiceinterfacesarethetaskrequirementsoftheapplication,thecapabilitiesandlimitationsofthetechnology,andthecharacteristicsoftheuserpopulation.
Thispaperdiscussedhowthesefactorsinuencevoiceuserinterfacedesignandthendescribescomponentsofvoiceuserinterfacesthatcanbeusedtofacilitateefcientandeffectivehuman-computervoice-basedinteractions.
6ContributionofThisStudyThisstudyintendstodemonstratewhethervariationsinvocalcharacteristicsofVUIinuenceuserexperienceandperception.
Indoingso,thisstudywillofferboththe-oreticalandpracticalinsightsformanufacturers,designers,andcommunicationscholarswhoareinterestedintheroleofVUIinpromotingadoptionforemergingsmartcommunicationmediasuchassmartphonesandsmartwatches.
Acknowledgment.
ThisresearchwassupportedbytheMinistryofEducation,SouthKorea,undertheBrainKorea21PlusProject(GrantNo.
10Z20130000013).
References1.
Addington,D.
W.
:Theeffectofvocalvariationsonratingsofsourcecredibility.
SpeechMonogr.
38,242–247(1971)2.
Sundar,S.
S.
,Nass,C.
:Sourceorientationinhuman-computerinteractionprogrammer,networker,orindependentsocialactor.
Commun.
Res.
27(6),683–703(2000)3.
Morishima,Y.
,Nass,C.
,Bennett,C.
,Lee,K.
M.
:Effectsof'gender'ofcomputer-generatedspeechoncredibility.
TechnicalreportofIEICETL2001-16,31(8),pp.
557-562(2001)4.
Reeves,B.
,Nass,C.
:HowPeopleTreatComputers,Television,andNewMediaLikeRealPeopleandPlaces.
CSLIPublicationsandCambridgeUniversityPress,NewYork(1996)5.
Ko,S.
J.
,Judd,C.
M.
,Blair,I.
V.
:Whatthevoicereveals:within-andbetween-categorystereotypingonthebasisofvoice.
Pers.
Soc.
Psychol.
Bull.
32(6),806–819(2006)6.
Leigh,T.
W.
,Summers,J.
O.
:Aninitialevaluationofindustrialbuyers'impressionsofsalespersons'nonverbalcues.
J.
Pers.
SellingSalesManage.
22(1),41–53(2002)7.
Vacher,M.
,Fleury,A.
,Portet,F.
,Serignat,J.
F.
,Noury,N.
:Completesoundandspeechrecognitionsystemforhealthsmarthomes:applicationtotherecognitionofactivitiesofdailyliving.
In:NewDevelopmentsinBiomedicalEngineering,pp.
645–673(2010)8.
Rougui,J.
E.
,Istrate,D.
,Souidene,W.
:Audiosoundeventidenticationfordistresssituationsandcontextawareness.
In:AnnualoftheIEEEInternationalConferenceonEngineeringinMedicineandBiologySociety,EMBC2009,pp.
3501–3504.
IEEE(2009)9.
Lines,L.
,Hone,K.
S.
:Multiplevoices,multiplechoices:olderadults'evaluationofspeechoutputtosupportindependentliving.
Gerontechnology5(2),78–91(2006)10.
Gdde,F.
,Mller,S.
,Engelbrecht,K.
P.
,Kühnel,C.
,Schleicher,R.
,Naumann,A.
,Wolters,M.
:Studyofaspeech-basedsmarthomesystemwitholderusers.
In:InternationalWorkshoponIntelligentUserInterfacesforAmbientAssistedLiving,pp.
17–22(2008)290J.
JeongandD.
-H.
Shin11.
Hamill,M.
,Young,V.
,Boger,J.
,Mihailidis,A.
:Developmentofanautomatedspeechrecognitioninterfaceforpersonalemergencyresponsesystems.
J.
NeuroEngineeringRehabil.
6(1),26(2009)12.
López-Cózar,R.
,Callejas,Z.
:Multimodaldialogueforambientintelligenceandsmartenvironments.
In:Nakashima,H.
,Aghajan,H.
,Augusto,J.
C.
(eds.
)Handbookofambientintelligenceandsmartenvironments,pp.
559–579.
Springer,Heidelberg(2010)13.
Stevens,C.
,Lees,N.
,Vonwiller,J.
,Burnham,D.
:On-lineexperimentalmethodstoevaluatetext-to-speech(TTS)synthesis:effectsofvoicegenderandsignalqualityonintelligibility,naturalnessandpreference.
Comput.
SpeechLang.
19(2),129–146(2005)14.
Lines,L.
,Hone,K.
S.
:Multiplevoices,multiplechoices:olderadults'evaluationofspeechoutputtosupportindependentliving.
Gerontechnology5(2),78–91(2006)15.
Mullennix,J.
W.
,Stern,S.
E.
,Wilson,S.
J.
,Dyson,C.
L.
:Socialperceptionofmaleandfemalecomputersynthesizedspeech.
Comput.
Hum.
Behav.
19(4),407–424(2003)16.
Mairesse,F.
,Walker,M.
A.
,Mehl,M.
R.
,Moore,R.
K.
:Usinglinguisticcuesfortheautomaticrecognitionofpersonalityinconversationandtext.
J.
Artif.
Intell.
Res.
30,457–500(2007)17.
Roe,D.
B.
,Wilpon,J.
G.
(eds.
):VoiceCommunicationBetweenHumansandMachines.
NationalAcademiesPress,Washington,DC(1994)18.
Brann,M.
,Himes,K.
L.
:Perceivedcredibilityofmaleversusfemaletelevisionnewscasters.
Commun.
Res.
Rep.
27(3),243–252(2010)19.
Niculescu,A.
,VanDijk,B.
,Nijholt,A.
,See,S.
L.
:Theinuenceofvoicepitchontheevaluationofasocialrobotreceptionist.
In:2011InternationalConferenceonUserScienceandEngineering(i-USEr),pp.
18–23.
IEEE(2011)20.
Nass,C.
,Lee,K.
M.
:Doescomputer-synthesizedspeechmanifestpersonalityExperimentaltestsofrecognition,similarity-attraction,andconsistency-attraction.
J.
Exp.
Psychol.
Appl.
7(3),171(2001)It'snotWhatItSpeaks,butIt'sHowItSpeaks291