2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks

2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks

ID:40705879

大小:405.29 KB

页数:11页

时间:2019-08-06

2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks_第1页
2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks_第2页
2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks_第3页
2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks_第4页
2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks_第5页
资源描述:

《2016-EMNLP-Speculation and Negation Scope Detection via Convolutional Neural Networks》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、SpeculationandNegationScopeDetectionviaConvolutionalNeuralNetworks11112ZhongQian,PeifengLi,QiaomingZhu,GuodongZhou,ZhunchenLuoandWei2Luo1SchoolofComputerScienceandTechnology,SoochowUniversity,Suzhou,215006,China2ChinaDefenseScienceandTechnologyInformationCenter,Bei

2、jing,100142,Chinaqianzhongqz@163.com,{pfli,qmzhu,gdzhou}@suda.edu.cn,zhunchenluo@gmail.com,htqxjj@126.comsentencesforexamples:Abstract(S1)Thedoctorswarnthatsmoking[mayharmourlungs].Speculationandnegationareimportantin-(S2)Hedoes[notlikeplayingfootball]butformationt

3、oidentifytextfactuality.Inthis1paper,weproposeaConvolutionalNeurallikesswimming.Network(CNN)-basedmodelwithprobabil-InsentenceS1,thespeculativecue“may”go-isticweightedaveragepoolingtoaddressvernsthescope“mayharmourlungs”,whilethespeculationandnegationscopedetection

4、.Innegativecue“not”governsthescope“notlikeparticular,ourCNN-basedmodelextractsplayingfootball”insentenceS2.thosemeaningfulfeaturesfromvarioussyn-tacticpathsbetweenthecuesandthecandi-Previousworkhaveachievedquitesuccessondatetokensinbothconstituencyanddepen-cueident

5、ification(e.g.,withF1-scoreof86.79dencyparsetrees.EvaluationonBioScopeforspeculativecuedetectioninTangetal.showsthatourCNN-basedmodelsignificant-(2010)).Incomparison,speculationandnegationlyoutperformsthestate-of-the-artsystemsonscopedetectionisstillachallengedueto

6、itsin-Abstracts,asub-corpusinBioScope,andherentdifficultiesandthoseupstreamerrors.InachievescomparableperformancesonClini-thispaper,wefocusonscopedetection.PreviouscalRecords,anothersub-corpusinBioScope.workonscopedetectioncanbeclassifiedintoheuristicrulesbasedmeth

7、ods(e.g.,Özgüretal.,1Introduction2009;Øvrelidetal.,2010),machinelearningFactualinformationiscriticaltounderstandabasedmethods(e.g.,Tangetal.,2010;Zouetal.,sentenceoradocumentinmosttypicalNLPap-2013),andhybridapproacheswhichintegrateplications.Speculationandnegation

8、extractionempiricalmodelswithmanualrules(Velldalethasbeendrawingmoreandmoreattentionsinal.,2012).recentyearsduetoitsimportanceindistinguish-Diffe

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。