confidence measures, dialog modeling and user interface外语英文电子书

confidence measures, dialog modeling and user interface外语英文电子书

ID:7950644

大小:5.80 MB

页数:35页

时间:2018-03-02

confidence measures, dialog modeling and user interface外语英文电子书_第1页
confidence measures, dialog modeling and user interface外语英文电子书_第2页
confidence measures, dialog modeling and user interface外语英文电子书_第3页
confidence measures, dialog modeling and user interface外语英文电子书_第4页
confidence measures, dialog modeling and user interface外语英文电子书_第5页
资源描述:

《confidence measures, dialog modeling and user interface外语英文电子书》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库

1、CHAPTER3CONFIDENCEMEASURES,DIALOGMODELINGANDUSERINTERFACESUMMARYTheprevioustwochaptersreviewed,respectively,themainsourcesofvariability/distortionandthecurrentstateoftheartofenvironment-independent/adaptivespeechrecognitiontechnology.Inthischapterwediscusswhatarethecriticaltechnologies

2、/componentsthatmakespeechrecognitionusable.Whendesigninganapplication,issuessuchashowtoassignaconfidencescoretoanhypothesis,howtodealwithextraneousspeechandout-ofvocabularywords,andhowtoprovideuserfeedbackanddesignefficienterrorcorrectiontechnologyisofprimeimportance.Moreover,manyappli

3、cationsnecessitatethedevelopmentofadialogmodulewhichmodelswhatisgoingonintheinteraction,capturesthecontextofthecommunication(whathasbeensaidbefore)andhelpstheusercompletehisrequest(ifneeded)andtoperformactions.Thischapteremphasizesthesetechnologiesandprovidessomedesignrecommendationson

4、howtodesignprompts,generateuserfeedbackanddevelopspeechuserinterfaces.75Chapter3ConfidenceMeasures,DialogModelingandUserInterface3.1.INTRODUCTIONInthelastdecade,theperformanceofspeechrecognitionsystemshasreachedanacceptablerange,makingpossiblemanyapplicationssuchasvoicedialing,informat

5、ionretrievaloverthephoneanddictation.Thesesystemsperformreasonablywellwhentheuseriscooperativeandcarefullyusesthesystem’svocabulary.However,whentheuser’sspeechcontainsoutofvocabularywordsordeviatesfromthetaskgrammaracceptedbythesystem,recognitionaccuracydegradessharply.Inparticular,whe

6、nspeakersarenoviceusersofthesystem,hesitations,Out-Of-Vocabulary(OOV)wordsandfillersoundsarebecomingaproblemforspeechrecognition.Inpracticalapplications,itiscommonthatabout20to30%oftheuserutterancesareoutofthegrammarspecifications.Ausablesystemshouldbeabletoignoreextraneousspeechandnoi

7、seandtodeterminehowconfidentitisinthespeechthathasbeenrecognized.Toextractpartialinformationfromaninpututteranceword-spottingalgorithmshavebeendeveloped.Thesealgorithmshavebeenappliedtoanumberofproblemsthatcansufferfromunexpectedspeechinput.Afterpartialinformationhasbeenextracted,iti

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。