entiti data manag okkam explor discov process view process spaceship entiti lifecycl manag okkam abstract paper examin special requir lifecycl manag entiti context entiti manag system semant web studi requir respect creat modifi entiti manag evolut time present issu aris access control model need manag larg distribut repositori entiti final discuss research direct offer solut problem give overview techniqu method relev solut direct autom gener semant annot tool cluster techniqu natur languag overview semant issu text mine clean answer dirti databas probabilist approach detect duplic tupl real world entiti import task data integr clean techniqu exist identifi tupl merg elimin duplic difficult task reli ad hoc manual solut propos complementari approach permit declar queri answer duplic data duplic probabl clean databas rewrit queri databas duplic return answer probabl answer clean databas rewritten queri sensit semant duplic user understand queri answer present clean databas semant adopt independ probabl produc effect exploit queri answer absenc extern knowledg associ databas tupl probabl offer techniqu base tupl summari autom task experiment studi perform rewritten queri studi show rewrit introduc signific overhead queri execut time work context conquer project univers toronto focus effici manag inconsist dirti databas improv build architectur legaci fc softwar system reduc build time precompil evolv larg softwar inform dtheoret softwar cluster limbo scalabl cluster categor data abstract cluster problem great practic import numer applic problem cluster challeng data categor inher distanc measur data valu introduc limbo scalabl hierarch categor cluster algorithm build inform bottleneck ib framework quantifi relev inform preserv cluster hierarch algorithm limbo advantag produc cluster size singl execut ib framework defin distanc measur categor tupl present distanc measur categor attribut valu show limbo algorithm cluster tupl valu limbo handl larg data set produc memori bound summari model data present experiment evalu limbo studi cluster qualiti compar categor cluster algorithm limbo support trade effici term space time qualiti quantifi trade demonstr limbo substanti improv effici neglig decreas qualiti inform dtheoret tool mine databas structur larg data set data design character process arriv design maxim inform content piec data equival minim redund inform content redund measur respect prescrib model data model express set constraint work problem data redesign environ prescrib model unknown incomplet specif problem find structur clue instanc data instanc error miss valu duplic record propos set inform theoret tool find structur summari character inform content data ultim data design provid algorithm creat summari larg categor data set studi summari specif physic design task rank function depend base data redund show rank physic data design tool find good vertic decomposit relat decomposit improv inform content design present evalu approach real data set kanata adapt evolut data share system categor cluster schema discoveri techniqu manag relat schema assum schema adequ model data owev practic semant data evolv time schema data structur constraint updat reflect common exampl includ overload tabl store fact type order tabl origin design store servic order store product order compani expand scope busi similarli semant typic repres constraint evolv data share origin semant full semant captur origin legaci design long term research goal find techniqu discov schema fit data work take initi step goal specif examin benefit categor cluster discov group tupl share similar structur characterist softwar cluster base inform loss minim schema discoveri structur data distinguish unstructur data presenc schema describ logic structur semant data schema mean understand queri underli data schema permit sophist structur queri schema data system assum schema predefin accur reflect data assumpt valid network databas data origin sourc valid legaci databas semant data evolv time result queri task depend structur queri includ data integr schema map effect paper problem discov schema data focu discov properti data exploit queri transform data final briefli suitabl mine approach task schema discoveri schema manag clio manag system heterogen data coupl tradit databas manag engin addit tool manag schema model data map schema articl provid overview clio place solut context rich research literatur data integr transform clio result collabor univers toronto ibm almaden research center address foundat system issu relat heterogen data schema integr manag revers engin meet data analysi demonstr data manag techniqu line analyt process olap enhanc sophist rang softwar revers engin tool comprehens examin similar differ task olap techniqu meet fail meet revers engin revers engin recast data analysi permit seamless integr technolog extend multidimension data model manag dynam chang dimens data aggreg case studi apach web server show solut permit integr view data rang low level program analysi inform abstract aggreg inform high level abstract provid human visual tool directli revers engin tool data mine techniqu 