资源描述:
《CDAS A Crowdsourcing Data Analytics System CDAS一个众包数据分析系统》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、CDAS:ACrowdsourcingDataAnalyticsSystemXuanLiu†,MeiyuLu†,BengChinOoi†,YanyanShen†,SaiWu§,MeihuiZhang††SchoolofComputing,NationalUniversityofSingapore,Singapore§CollegeofComputerScience,ZhejiangUniversity,Hangzhou,P.R.China†{liuxuan,lumeiyu,ooibc,shenyanyan,mhzhang}@comp.nu
2、s.edu.sg,§wusai@zju.edu.cnABSTRACTworkerHumanSomecomplexproblems,suchasimagetaggingandnaturallan-IntelligentTaskworkerguageprocessing,areverychallengingforcomputers,whereevenstate-of-the-arttechnologyisyetabletoprovidesatisfactoryaccu-racy.Therefore,ratherthanrelyingsolel
3、yondevelopingnewandworkerusercrowdsourcingjobbetteralgorithmstohandlesuchtasks,welooktothecrowdsourc-applicationHumaningsolution–employinghumanparticipation–tomakegoodtheIntelligentTaskworkershortfallincurrenttechnology.Crowdsourcingisagoodsupple-menttomanycomputertasks.A
4、complexjobmaybedividedintocomputerjobworkercomputer-orientedtasksandhuman-orientedtasks,whicharethenassignedtomachinesandhumansrespectively.Figure1:CrowdsourcingApplicationToleveragethepowerofcrowdsourcing,wedesignandimple-Yahoo!Answers,whereuserssubmitandanswerquestions.
5、InmentaCrowdsourcingDataAnalyticsSystem,CDAS.CDASisaWeb2.0sites,mostofthecontentsarecreatedbyindividualusers,frameworkdesignedtosupportthedeploymentofvariouscrowd-notserviceproviders.Crowdsourcingisthedrivingforceofthesesourcingapplications.ThecorepartofCDASisaquality-sen
6、sitivewebsites.Tofacilitatethedevelopmentofcrowdsourcingappli-answeringmodel,whichguidesthecrowdsourcingenginetopro-1cations,AmazonprovidestheMechanicalTurk(AMT)platform.cessandmonitorthehumantasks.Inthispaper,weintroducetheComputerprogrammerscanexploitAMT’sAPItopublishjo
7、bsforprinciplesofourquality-sensitivemodel.Tosatisfyuserrequiredhumanworkers,whoaregoodatsomecomplexjobs,suchasim-accuracy,themodelguidesthecrowdsourcingqueryenginefortheagetaggingandnaturallanguageprocessing.Thecollectiveintel-designandprocessingofthecorrespondingcrowdso
8、urcingjobs.ligencehelpssolvemanycomputationallydifficulttasks,therebyItprovidesanestimatedaccurac