One Sense Per Discourse

One Sense Per Discourse

ID:40086700

大小:455.10 KB

页数:5页

时间:2019-07-20

One Sense Per Discourse_第1页
One Sense Per Discourse_第2页
One Sense Per Discourse_第3页
One Sense Per Discourse_第4页
One Sense Per Discourse_第5页
资源描述:

《One Sense Per Discourse》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、OneSensePerDiscourseWilliamA.GaleKennethW.ChurchDavidYarowskyAT&TBellLaboratories600MountainAvenueMurrayHillNJ07974-06361.ABSTRACToftenbeusedinlieuofhand-labeling.Forexample,con-siderthepolysemouswordsentence,whichhastwoma-Itiswell-knownthattherearepolysemouswordslikesentencewhose"meaning"or"sense

2、"dependsonthejorsenses:(1)ajudicialsentence,and(2),asyntacticsentence.Wecancollectanumberofsense(1)examplescontextofuse.Wehaverecentlyreportedontwonewbyextractinginstancesthataretranslatedaspeine,andword-sensedisambiguationsystems,onetrainedonbilin-wecancollectanumberofsense(2)examplesbyextract-gu

3、almaterial(theCanadianHansards)andtheotheringinstancesthataretranslatedasphrase.Inthisway,trainedonmonolingualmaterial(Roget'sThesaurusandwehavebeenabletoacquireaconsiderableamountofGrolier'sEncyclopedia).Asthisworkwasnearingcom-testingandtrainingmaterialfordevelopingandtestingpletion,weobservedav

4、erystrongdiscourseeffect.Thatourdisambiguationalgorithms.is,ifapolysemouswordsuchassentenceappearstwoormoretimesinawell-writtendiscourse,itisextremelyTheuseofbilingualmaterialsfordiscriminationdeci-likelythattheywillallsharethesamesense.ThispapersionsinmachinetranlationhasbeendiscussedbyBrowndescr

5、ibesanexperimentwhichconfirmedthishypothesisandothers(1991),andbyDagan,Itai,andSchwallandfoundthatthetendencytosharesenseinthesame(1991).Theuseofbilingualmaterialsforanessentiallydiscourseisextremelystrong(98%).Thisresultcanbemonolingualpurpose,sensedisambiguation,issimilarinusedasanadditionalsour

6、ceofconstraintforimprovingmethod,butdiffersinpurpose.theperformanceoftheword-sensedisambiguationalgo-rithm.Inaddition,itcouldalsobeusedtohelpevaluate2.2.BayesianDiscriminationdisambiguationalgorithmsthatdidnotmakeuseofthediscourseconstraint.SurprisinglygoodresultscanbeachievedusingBayesiandiscrimi

7、nationmethodswhichhavebeenusedverysuc-cessfullyinmanyotherapplications,especiallyauthor2.OURPREVIOUSWORKONidentification(MostellerandWallace,1964)andinfor-WORD-SENSEDISAMBIGUATIONmationretrieval(IR)(Salton,1989,s

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。