-1- Design and Evaluation of the Asynchronous Voice Meeting System AVM54-第3页
上亿文档资料，等你来发现
-1- Design and Evaluation of the Asynchronous Voice Meeting System AVM54-3
messagesisinFigure5.Thel；ViaVoice98showed85.2%wor；Theaverageoftheresultoft；Table3Analysisofthemessa；VM；Totalcountofusingthesyst；Numberofthemessages；Numberofthecharactersofa；(excerp
　messagesisinFigure5.ThelengthsofthemessagesonAVMwerecountedusingthemanualtranscriptionofvoicemessagesaftertheexperiment.InregardstotheutterancesonBBS,thenumberofcharacterswerecountedexcludingthepartialexcerption.ViaVoice98showed85.2%wordrecognitionrateforcorrectnessand82.5%foraccuracy.TheaverageoftheresultofthequestionnaireisshowninTable4(withafive-gradesystem,including“donotthinksoatall”asone,and“thinksoverymuch”asfive.)Thequestions,(k)and(l),wereaskedonlytotheAVMusers.Table3AnalysisofthemessagesonAVMandBBS.AVMTotalcountofusingthesystemNumberofthemessagesNumberofthecharactersofallmessages(excerptionsareexcluded)AveragenumberofcharactersofamessageTotaltimeofmessages(sec)Averagetimeofamessage(sec)NumberoftheoverlappedmessagesNumberofthenon-overlappedmessages.59BBS.7----Fig.5LengthofthemessagesonAVMandBBS.Table4Resultofquestionnaire.Items(b)Iwasabletoadequatelyspeakmyopinions(c)Activediscussionwasabletobeconducted(d)ThediscussionproceededinanaturalmannerAVM4.03.84.0BBS4.43.44.0(e)Themoodwasverysimilartowhenpeopleactuallycongregatetoconductdiscussions(f)Itwaseasytoexpoundtheprosandcons(g)Easytoinputmessages(h)Easytograspthecontentofmessages(i)Easytoviewthediscussionsoftheentireconference(j)Theuserinterfacewassimpletounderstand(k)Iusedthetextdisplaymorethanthevoiceplay(l)Usingthetext,includingthemisrecognitions,wasvaluable2.24.43.43.43.43.84.04.62.83.84.24.03.84.0--6.3DiscussionWeshalldiscusstheresultinviewofthefollowinghypotheses:(1)AVMrealizedanasynchronousvoicemeetinglikespokendialogs.(2)AVMmadeuseoftheprogressiveprocessofspokenwords.(3)BISPfunctionworkedwellandcontributedtosmoothconversation.(4)Thespeechrecognitionresultincludingmisrecognitionswasefficientlyused.(5)AVMfunctionedsufficientlyasanalternativetoBBS.WewillfirstdiscussHypothesis1.WithAVM,thesubjectsmostlyspokewithconversationalandcasualexpressionssuchas“Well,talkingaboutsoccernow…about20teamsinaleague.Andtomake1000players,50foroneteam.Therearen’tsomanyonateam,arethere?”Onthecontrary,withBBS,theparticipantsmainlyutilizedwrittenlanguage,forexample,“Inmyinterpretation,asothersaresaying,thereshouldbemorethan84players()intheprofessionalbaseballleague,includingthefarmteams.However,thereisalsoalimitforthenumberofregisteredplayersforoneteam,so…”Therelationsamongtheteammemberscannotcausesuchadifferencebecauseallthesubjectskneweachotherwell.Accordingly,theresultsupportsHypothesis1.Hypothesis2predictstheeliminationofsharedinformation,whichcausesthefrequentutilizationofsimpleandfragmentedexpressionsasstatedintheSection2.Severalresultssupportedthishypothesis,includingthefactthatthetotallengthofmessageswithAVMwas48%ofthatofBBS(Table3),thatthelengthofmostmessageswererelativelyshort(Figure5),andthatmanyutterancesdependedonthecontextsuchas“Well,it’sclose,asyousaid.”Hypothesis3wasonlypartiallysupported.Manyoftheoverlappingutterances,whichare17%ofthetotalspeech,werefragmentedbecauseofunsuccessfuldetectionofthespeechsegment.Thisalsocausedthelowratingofquestion(g)inTable4,indicatinginsufficientprocessingofoverlappingspeeches.However,thehighratingofquestions(a)to(d)inTable4suggeststhatsmoothconversationforthisexperimentwasrealized.Hypothesis4wassupportedbytheresultsofthequestionnaire,includinghighevaluationofquestions(k),(l)inTable4,subjects’commentssuchas,‘Iconfirmedthecontextroughlybythetextandminutelybythereplay,’and,‘Thetextinformationsavedtimefromlisteningtovariousutterances.’Hypothesis5wassupportedbythefactsthatstatedinTable4,AVMwasnotratedfarlowerthanBBS,andthatbothteamsreachedacceptableanswers.However,ratingofthequestion(e)forAVMwasaslowasBBS,whichindicatesthatAVMcannotreplaceface-to-facediscussionyet.Accordingly,theexperimentresultssupporthypothesis1,2,4,and5completely,andhypothesis3partially.7.ConclusionInthisreport,wedesignedandevaluatedAVM,anasynchronous,non-realtime,andinteractivevoicemeetingsystem.WewereabletoconfirmthatAVMappliestheadvantagesofvoicemessaginginthatitishighlyexpressive,easytotalk,anddoesnotimpairthemeritsofasynchronouscommunicationofelectronicmail,bulletinboardsystems,etc.Inthefuture,weintendtoadvancetheperformanceofspeechrecognitionintheserverparticularlyforspokenlanguage.Improvementsforsmootherconversationarealsorequired,includingavoidingredundantreplayingofutterancesalreadyheard,andpreventingthefragmentingofmessagesinthemiddleofaword.Inaddition,large-scaleoperatingtestswithvarioususers,suchaspeopleunfamiliartothekeyboard,arenecessary.WeplantofurtherinvestigatetherealizationofanAVMapplicationontelephonyservicesandhandhelddevices,toattractawiderrangeofusers.(Note1):SynchronizedMultimedia,(Note2):ExtensibleMarkupLanguage,(Note3):HypertextTransferProtocolHTTP/1.1,RFC2068.(Note4):References[1]M.J.Hirayama,T.Sugahara,Z.Peng,J.Yamazaki:InteractivelisteningtostructuredspeechcontentontheInternet,ProceedingsofICSLP'98,pp.,Dec.1998.[2]M.Okada:Themumblingcomputer,KyoritsuShuppan,Tokyo,1995(inJapanese).[3]T.Nishimoto,Y.Niimi:Anasynchronousvoicemessagingsystem,IEICETechnicalReport,MVE97-98,pp.39-46,Feb.1998(inJapanese).[4]M.EnomotoandS.Tutiya:Overlappingandit’sstatisticalanalysisintheJapanesemaptaskdialoguecorpus,IPSJSIGNotes,99-SLP-29-25,pp.145-150,Dec.1999(inJapanese).[5]Y.Kawaguchi,S.Tutiya:Onthoseaberrationsfromtheturn-takingruleswhichcanbeaccountedforbyconversationalimplicature,IPSJSIGNotes,99-SLP-29-26,pp.151--156,Dec.1999(inJapanese).[6]H.Nishi,K.Gomi,J.Kojima:Stochasticspeech-enddetectionforman-machinespeechcommunication,TransactionsoftheIEICE,D,Vol.J70-D,No.11,pp.,Nov.1987(InJapanese).[7]C.Kougo,J.Yamauchi:Anexperimentalansweringmachinewithbackchannelresponses,The8thannualmeetingofJapaneseCognitiveScienceSociety,pp.72-73,Jul.1991(InJapanese).[8]N.Ward:Usingprosodiccluestodecidewhentoproduceback-channelutterances,ProceedingsofICSLP'96,pp.,Oct.1996.[9]H.Kikuchi,Y.Sugita,K.Shirai:Analysisofinfluenceoftimeconstraintinnaturaldialogue,JSAISIGNotes,SIG-SLUD-9702-5,pp.31-36,Oct.1997(InJapanese).[10]Y.Horiuchi,H.Koiso,S.TutiyaandA.Ichikawa:Someutterance-finalsyntacticandprosodiccharacteristicsrelevanttothecontrolofthespeakershiftphenomenainspontaneousspokendialogues,IPSJSIGNotes,96-SLP-10-9,pp.45--50,Feb.1996(InJapanese).[11]Y.Okato,KeijiKato,M.YamamotoandS.Itahashi:Evaluationofthedialogsystemswithback-channelfeedbacks,JSAISIGNotes,SIG-SLUD-9804-2,pp.7-12,Feb.1999(inJapanese).[12]J.Hirasawa,N.Miyazaki,M.Nakano,T.Kawabata:Implementationofcoordinativenoddingbehavioronspokendialogsystems,ProceedingsofICSLP'98,pp.,Dec.1998.[13]T.Nishimoto,H.Yuki,T.KawaharaandY.Niimi:Anasynchronousvirtualmeetingsystemforbi-directionalspeechdialog,ProceedingsofEurospeech'99,pp.,Sep.1999.ProfilesTakuyaNishimotoHereceivedB.E.andM.E.fromWasedaUniversity,Japanin1993and1995,respectively.HeisaResearchAssociateatDepartmentofElectronicsandInformationScience,KyotoInstituteofTechnology,Japan.Hisresearchinterestsincludespokendialoguesystemsandhuman-machineinterfaces.HeisamemberofInformationProcessingSocietyofJapan,theAcousticalSocietyofJapan,theJapanSocietyforArtificialIntelligence,andHumanInterfaceSociety.HidehiroYukiHereceivedB.E.andM.E.fromKyotoInstituteofTechnology,Japanin1998and2000,respectively.TakehikoKawaharaHereceivedB.E.andM.E.fromKyotoInstituteofTechnology,Japanin1999and2001,respectively.MasahiroArakiHereceivedM.E.ininformationscienceandD.Eng.fromKyotoUniversity,Kyoto,Japan,in1990and1998,respectively.Technology.HeisanAssociateProfessoratDepartmentofElectronicsandInformationScience,KyotoInstituteofHisresearchinterestsincludespokendialoguesystemsandnaturallanguageunderstanding.HeisamemberofIPSJ,JSAI,ANLP,andACL.YasuhisaNiimiHereceivedtheB.E.,theM.E.andthePh.D.degreesfromKyotoUniversity,Japaninand1969,respectively.FromhewasworkingatKyotoUniversity.Since1970hehasbeenworkingattheDepartmentofElectronicsandInformationScienceofKyotoInstituteofTechnology,whereheisnowaprofessor.Hiscurrentinterestsincludespeechinformationprocessing,naturallanguageandartificialintelligence.Hepublishedthebool“SpeechRecognition”(inJapanese)in1979.Dr.NiimiisamemberoftheInformationProcessingSocietyofJapan,theAcousticalSocietyofJapan,andtheJapanSocietyforArtificialIntelligence.包含各类專业文献、文学作品欣赏、行业资料、外语学習资料、高等教育、幼儿教育、小学教育、生活休闲娱乐、应用写作文书、专业论文、各类資格考试、-1- Design and Evaluation of the Asynchronous Voice Meeting System AVM54等内容。　
　　【】　
您可在本站搜索以下内容：
赞助商链接
别人正在看什么？
贊助商链接下次自动登录&|
让卖家联系您
扫描我，即刻搜索
指针钳型表优质产品
共找到15条"
指针鉗型表"
您是不是要找:
所有经营模式
执照已审核
匼并相同供应商
没有找到合适的指针钳型表供應商？您可以
您可能感兴趣的
最新入驻厂家
最噺相关产品
正在被搜索
没找到想要的产品？
你昰不是要找
相关的产品信息？
发布询价单让供應商主动联系找到你
相关区域产品:
正在被查看：
指针钳型表产品参数信息由世界工厂网为您提供，您可以查看到指针钳型表相关的优质批發/供应及指针钳型表厂家/经销商信息，同时为您提供全面的指针钳型表出厂价格参考，最新嘚指针钳型表报价尽在世界工厂网
按字母分类: