文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Continuous Speech Recognition at
侵权投诉

Continuous Speech Recognition at

Abstract: This paper summarizes recent activities at LIMSI in multilingual speech recognition and its applications. While the main goal of speech recognition is to provide a transcription of the speech signal as a sequence of words, the same basic technolo

observedinthelinguisticproperties,thedifferencesinreadandspontaneousspeechwereanalyzed.Asaresult, llerwordsandbreathnoiseareexplicitlyrepresentedintheacousticandlanguagemodels.Compoundwordswereintroducedasameansofmodelingreducedpronunciationsforcommonwordsequences.Theworderrorforauto-matictranscriptionofcontinuousbroadcastnewsdataisaround20%.Thisperformanceseemssuf cientforpotentialapplicationsasdemonstratedbytheInformediaprojectatCMU.LIMSIisparticipatingintheLEOLIVEprojectpro-vidingthespeechtechnologyforautomaticindexationofFrenchandGermanbroadcasts.Relatedtechnologiessuchasspeakerorgenderidenti cationareappliedtoimproverecognitionperformance.

SPOKENLANGUAGESYSTEMSFORINFORMATIONRETRIEVAL

Spokenlanguagesystems(SLSs)aimtohelpauseraccomplishataskviainteractivedialog.Taskanddomainknowledgemustbeusedtode nethevocabularyandtheconceptsspeci ctotheapplicationinordertoconstructappropriateacoustic,languageandsemanticmodels.Modelizationofspontaneousspeecheffects,suchashesitations,falsestarts,andreparations,isparticularlyimportantforthesesystems.Incontrasttodictationandtranscriptiontaskswhereitisrelativelystraight-forwardtoselectarecognitionvocabularyfromlargewrittencorpora,forspokenlanguagesystemsthereusuallyarenoapplication-speci ctrainingdata(acousticortextual)available.Acommonlyadoptedapproachfordatacollectionistostartwithaninitialsystem(thatmayinvolveaWizardofOzcon guration)andtocollectasetofdatawhichcanbeusedtostartaniterativedevelopmentcycle.

Inadditiontoaspeaker-independent,continuousspeechrecognizer,aspokenlanguagedialogsystemalsoincludescomponentsfornaturalspokenlanguageunderstanding,dialogmanagement,historymanagement,databaseaccess,responsegeneration,andspeechsynthesis.Thespeechrecognizertransformstheinputsignalintothemostprobablewordsequence(oroptionallyawordgraph),andforwardsittotherule-basednaturallanguageunderstandingcompo-nent,whichgeneratesasemanticframe.Amixed-initiativedialogmanagerpromptstheusertosupplyanymissinginformationneededfordatabaseaccessandthengeneratesadatabasequery.Theretrievedinformationistransformedintonaturallanguagebytheresponsegenerator(takingintoaccountthedialogcontext)andpresentedtotheuser.Synthesisbywaveformconcatenationisusedtoensurehighqualityspeechoutput,wheredictionaryunitsareputtogetheraccordingtothegeneratedtextstring.Itisbecomingincreasingclearthatdialogmanagementandresponsegenerationplayanimportantroleinsystemdesignandusersatisfaction.

AtLIMSIprototypesystemstoprovidevocalaccesstotraintravelinformationhavebeendevelopedinthecontextofseveralEuropeanprojects,ESPRIT-MASK(3)andLERAILTEL,ARISE(8).Developmentofsystemsformoregeneraltouristinformation(AUPELF)andcontrolofhouseholdappliances(TideHOME)arealsounderway.Thesesystemsarebeingtestedin eldtrialswithnaiveusers.

REFERENCES

1.Adda,G.,Adda-Decker,M.,Gauvain,J.L.andLamel,L.,“TextNormalizationandSpeechRecognitioninFrench,”ProceedingsoftheEuropeanConferenceonSpeechTechnology,EuroSpeech,Rhodes,Greece,1997.

2.Gauvain,J.L.,Adda,G.,Lamel,L.,andAdda-Decker,M.,“Transcribingbroadcastnews:TheLIMSINov96Hub4System,”ProceedingsofARPASpokenLanguageTechnologyWorkshop,Chantilly,Virginia,56–63,1997.

3.Gauvain,J.L.,Bennacef,S.K.,Devillers,L.,Lamel,L.F.andRosset,S.,“TheSpokenLanguageComponentoftheMaskKiosk,”inHumanComfortandSecurityofInformationSystems,AdvancedInterfacesfortheInformationSociety,editorsK.C.VargheseS.P eger,SpringerVerlag,1997,pp93–103.

4.Gauvain,J.L.,andLamel,L.,“LargeVocabularyContinuousSpeechRecognition:fromLaboratorySystemstowardsReal-WorldApplications,”IEICEJournalJ79-D-II,2005–2021,(1996).

5.Gauvain,J.L.,Lamel,L.,Adda,G.,andAdda-Decker,M.,“Speaker-IndependentContinuousSpeechDictation,”SpeechCommunication15,21–37,(1994).

6.Gauvain,J.L.,Lamel,L.,Adda,G.,andAdda-Decker,M.,“TranscriptionofBroadcastNews,”ProceedingsoftheEuropeanConferenceonSpeechTechnology,EuroSpeech,Rhodes,Greece,1997.

7.Lamel,L.F.andAdda,G.,“OnDesigningPronunciationLexiconsforLargeVocabulary,ContinuousSpeechRecognition,”Proc.ICSLP’96,Philadelphia,PA(1996)

8.Lamel,L.F.,Gauvain,J.L.,Bennacef,S.K.,Devillers,L.,Foukia,S.,Gangolf,J.J.,Rosset,S.,“FieldTrialsofaTelephoneServiceforRailTravelInformation,”toappearinSpeechCommunication(1998).

9.Young,S.,Adda-Decker,M.,Aubert,X.,Dugast,C.,Gauvain,J.L.,Kershaw,D.J.,Lamel,L.,vanLeeuwen,D.A.,Pye,D.,Robinson,A.J.,Steeneken,H.J.M.,Woodland,P.C.,“MultilingualLargeVocabularySpeechRecognition:TheEuropeanSQALEProject,”ComputerSpeechandLanguage11(1),73-89(1997).

上一页第2页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email