DetectioniofiTruthiDiscoveryi nBigiDataiSocial iMediaiSensingiApplications

Abstract

With ithe irapid igrowth iof ionline isocial imedia iand iubiquitous iInternet iconnectivity, isocial isensing ihas iemerged ias ia inew icrowd isourcing iapplication iparadigm iof icollecting iobservations i(often icalled iclaims) iabout ithe iphysical ienvironment ifrom ihumans ior idevices ion itheir ibehalf. iA ifundamental iproblem iin isocial isensing iapplications iliesiin ieffectively iascertaining ithe icorrectness iof iclaims iand ithe ireliability iof idata isources iwithout iknowing ieither iof ithem ia ipriori, iwhich iis ireferred ito ias itruth idiscovery. iWhile isignificant iprogress ihas ibeen imade ito isolve ithe itruth idiscovery iproblem, isome iimportant ichallenges ihave inot ibeen iwell iaddressed iyet. iFirst, iexisting itruth idiscovery isolutions idid inot ifully isolve ithe idynamic itruth idiscovery iproblem iwhere ithe iground itruth iof iclaims ichanges iover itime. iSecond, imany icurrent isolutions iare inot iscalable ito ilarge-scale isocial isensing ievents ibecause iof ithe icentralized inature iof itheir itruth idiscovery ialgorithms. iThird, ithe iheterogeneity iand iunpredictability iof ithe isocial isensing idata itraffic ipose iadditional ichallenges ito ithe iresource iallocation iand isystem iresponsiveness. iIn ithis ipaper, iwe idevelop ia iScalable iand iRobust iTruth iDiscovery i(SRTD) ischeme ito iaddress ithe iabove ithree ichallenges. iIn iparticular, ithe iSRTD ischeme ijointly iquantifies iboth ithe ireliability iof isources iand ithe icredibility iof iclaims iusing ia iprincipled iapproach. iThe ievaluation iresults ion ithree ireal-world idata itraces i(i.e., iBoston iBombing, iParis iShooting iand iCollege iFootball) ishow ithat ithe iSSTD ischeme iis iscalable iand ioutperforms ithe istate-of-the- iart itruth idiscovery imethods iin iterms iof iboth ieffectiveness iand iefficiency.