Referring Expression Generation and Comprehension via Attributes

Referring Expression Generation and Comprehension via Attributes

ID:39716467

大小:2.15 MB

页数:9页

时间:2019-07-09

Referring Expression Generation and Comprehension via Attributes_第1页
Referring Expression Generation and Comprehension via Attributes_第2页
Referring Expression Generation and Comprehension via Attributes_第3页
Referring Expression Generation and Comprehension via Attributes_第4页
Referring Expression Generation and Comprehension via Attributes_第5页
资源描述:

《Referring Expression Generation and Comprehension via Attributes》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、ReferringExpressionGenerationandComprehensionviaAttributesJingyuLiu1,3LiangWang1,2,3Ming-HsuanYang41CenterforResearchonIntelligentPerceptionandComputing(CRIPAC),NationalLaboratoryofPatternRecognition(NLPR)2CenterforExcellenceinBrainScienceandIntelligenceTechnology(CEBSIT),InstituteofAutom

2、ation,ChineseAcademyofSciences(CASIA)3UniversityofChineseAcademyofSciences(UCAS)4UniversityofCalifornia,Merced{jingyu.liu,wangliang}@nlpr.ia.ac.cnmhyang@ucmerced.eduAbstractReferringexpressionisakindoflanguageexpressionthatusedforreferringtoparticularobjects.Tomaketheexpressionwithoutambi

3、guation,peopleoftenuseattributestodescribetheparticularobject.Inthispaper,weexploretheroleofattributesbyincorporatingthemintobothre-manwithredtiedonutwithholeferringexpressiongenerationandcomprehension.Wefirsttrainanattributelearningmodelfromvisualobjectsandtheirpaireddescriptions.Theninth

4、egenerationtask,wetakethelearnedattributesastheinputintothegenerationmodel,thustheexpressionsaregenerateddrivenbybothattributesandthepreviouswords.Forcomprehension,weembedthelearnedattributeswithvisualfeaturesandse-reddoubledeckerbusmostfrontelephantmanticsintothecommonspacemodel,thenthet

5、argetob-jectisretrievedbasedonitsrankingdistanceinthecommonFigure1.Someexamplesofourresults.Thefirstrowaretwospace.Experimentalresultsonthethreestandarddatasets,examplesofreferringexpressiongeneration.ThetargetobjectisRefCOCO,RefCOCO+,andRefCOCOgshowsignificantshownincontour.Thesecondroware

6、twoexamplesofreferringimprovementsoverthebaselinemodel,demonstratingthatexpressioncomprehension.Thegreenboxesaregroundtruthandourmethodiseffectiveforbothtasks.theblueboxesareretrievedboxes.naturallanguagegeneration.Ontheotherside,thetaskof1.Introductioncomprehensionrequiresthelistenertopr

7、ovehiscompre-Referringexpression[27]isaparticularkindofhu-hensionbypointingtothetargetobject’slocation,whichismanexpressionsthatfocusesoneffectivelydescribingthebasedonthequalityofthelistener’sinterpretationoftheex-uniqueobjectinsomeenvironments.Theinteractivepro-pr

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。