返回
published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew

published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew

ID：34508707

大小：111.30 KB

页数：5页

时间：2019-03-07

published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew_第1页
published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew_第2页
published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew_第3页
published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew_第4页
published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew_第5页
资源描述：

《published in proceedings of icassp-97 janus-iii speech-to-speech translation in multiple lanew》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、JANUS-III:SPEECH-TO-SPEECHTRANSLATIONINMULTIPLELANGUAGESAlonLavieAlexWaibelLoriLevinMichaelFinkeDonnaGatesMarsalGavaldaTorstenZeppenfeldPumingZhanInteractiveSystemsLaboratoriesCarnegieMellonUniversity,Pittsburgh,USA.UniversityofKarlsruhe,Karlsruhe,GermanyEmail:l

2、avie@cs.cmu.eduABSTRACTSpeechinSourceLanguageThispaperdescribesJANUS-III,ourmostrecentversionSpeechRecognizeroftheJANUSspeech-to-speechtranslationsystem.Wepresentanoverviewofthesystemandfocusonhowsys-N-bestListtemdesignfacilitatesspeechtranslationbetweenmultiplel

3、anguages,andallowsforeasyadaptationtonewsourceandtargetlanguages.WealsodescribeourmethodologyGLRTranslationModulePhoenixTranslationModuleforevaluationofend-to-endsystemperformancewithava-MorphAnalyzerGLR*ParserPhoenixParserrietyofsourceandtargetlanguages.Forsyste

4、mdevel-opmentandevaluation,wehaveexperimentedwithbothDiscourseInterlinguapush-to-talkaswellascross-talkrecordingconditions.ToProcessorFeatureStructureFrame-SlotStructuredate,oursystemhasachievedperformancelevelsofover80%acceptabletranslationsontranscribedinput,an

5、doverDiscourseMemory&70%acceptabletranslationsonspeechinputrecognizedCalendarGenKitGeneratorPhoenixGeneratorwitha75-90%wordaccuracy.Ourcurrentmajorresearchisconcentratedonenhancingthecapabilitiesofthesystemtodealwithinputinbroadandgeneraldomains.TextinTargetLangu

6、ageSpeechSythesizer1.INTRODUCTIONJANUS-IIisthemostrecentversionofJANUS,aspeech-SpeechinTargetLanguageto-speechtranslationsystem,designedtotranslatesponta-neousdialogsbetweenmultiplespeakers.JANUSisdevel-Figure1.TheJANUSSystemopedattheInteractiveSystemsLaboratorie

7、satCarnegieMellonUniversityandtheUniversityofKarlsruhe.Theformation.TheinterlinguarepresentationinoursystemcurrentsystemisdesignedfortheSchedulingdomain,inwasdesignedtocapturemeaningatthelevelofsuchSDUs.whichtwopartiesareparticipatinginanegotiationdialogEachseman

8、ticdialogunitisanalyzedintoaninterlinguainanattempttoscheduleameeting.representation.Forbothparsers,segmentationintoSDUsAcomponentdiagramofoursystemcanbeseenin

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。