suffix tree construct algorithm modern hardwar effici identif coupl entiti document collect twittermonitor trend detect twitter stream crowd paper introduc system call crowd protect user anonym worldwid web crowd name notion blend crowd oper group user larg geograph divers group crowd collect issu request behalf member web server unabl learn true sourc request equal origin member crowd collabor crowd member distinguish origin request member forward request behalf describ design implement secur perform scalabl system secur analysi introduc degre anonym import tool describ prove anonym properti earli onlin identif attent gather item social media transpar anonym thwart adversari algorithm transpar anonym thwart adversari algorithm identifi attribut describ spatial burst user gener content appear weblog wiki social network increas unpreced rate wealth inform produc individu geograph locat present challeng task intellig process paper introduc methodolog identifi notabl geograph focus event collect user gener inform heart propos lie effici algorithm identifi geograph focus inform burst attribut demograph factor identifi set descript keyword present result prototyp evalu algorithm blogscop larg scale social media wareh platform demonstr scalabl practic util propos run top multi terabyt text collect access cost dawar approach object retriev multipl sourc sourc object select retriev larg multi sourc data set fundament oper applic paper initi research effici sourc databas object select algorithm larg multi sourc data set specif order acquir number satisfi object minimum cost multipl databas queri engin determin access overhead individu data sourc overhead retriev object sourc possibl statist estim frequenc find satisfi object order determin object retriev data sourc adopt probabilist approach sourc select util cost structur dynam program model comput optim number object retriev data sourc structur valuabl asset monetari time relat cost access larg distribut databas present experiment evalu valid techniqu real world data set mrshare share multipl queri mapreduc larg scale data analysi li core modern enterpris scientif research emerg cloud comput analyt queri process infrastructur amazon ec directli map monetari mapreduc popular framework context cloud comput design serv long run queri job process batch mode take account job perform similar work opportun share principl share similar work reduc amount work lead reduc monetari charg incur util process infrastructur paper propos share framework tailor mapreduc framework mrshare transform batch queri batch execut effici merg job group evalu group singl queri base cost model mapreduc defin optim problem provid solut deriv optim group queri experi prototyp built top hadoop demonstr effect approach substanti save stream multipl aggreg phantom effici identif starter follow social media interact queri refin investig problem refin sql queri satisfi cardin constraint queri result applic answer problem face databas user formal problem queri refin propos framework support databas system introduc interact model refin incorpor user feedback captur user prefer techniqu design handl queri rang equal predic numer categor attribut present experiment evalu framework implement open sourc data manag demonstr feasibl practic util approach rank object base relationship fix associ text corpora enhanc addit metadata relat real world entiti document entiti discuss relationship typic obtain wide inform extract tool time interest associ typic hold entiti instanc corpu discuss hotel citi airlin fix associ entiti includ airlin oper flight citi hotel locat citi plethora applic necessit identif entiti match set keyword sampl queri find holidai packag pet friendli hotel locat histor live citi travel oper econom safe airlin keyword occur textual descript entiti actual hotel citi airlin answer queri exploit relationship entiti document keyword pet friendli occur document entiti hotel associ entiti hotel locat citi work focu class entiti packag finder queri outlin demonstr exist techniqu effici adapt solv problem result algorithm reli estim excess runtim storag overhead propos effici algorithm process queri larg corpora devis earli prune termin strategi presenc join aggreg execut entiti extract text depend estim analysi experiment evalu real synthet data demonstr effici scalabl approach metric function depend construct valid question measur attribut measur discuss formal theoret wai comput literatur term represent theori measur rare simpler wai foster applic practition construct valid start analysi construct attribut attempt measur ieee standard direct measur valid direct measur attribut involv metric depend attribut softwar engin attribut task simpl measur direct metric valid paper continu framework evalu propos metric appli bug count bug count captur small part mean attribut measur multidimension analys attribut promis mean captur qualiti attribut question analysi fragment run paper illustr breakdown attribut task interest attribut group studi join reorder join simul abstract introduc framework reorder join pipelin runtim databas system framework incorpor techniqu simul execut join pipelin random sampl statist summari simul techniqu provid accur runtim cardin estim altern execut path join pipelin estim util compar cost altern execut path dynam fashion reorder pipelin altern path found describ simul techniqu pipelin kind join oper present experiment evalu prototyp implement framework open sourc data manag result demonstr feasibl util approach present inform cascad blogospher curtain grapevin increment mainten length normal index approxim string match approxim string match problem receiv lot attent recent exist work inform retriev concentr varieti similar measur tf idf bm hmm specif tailor document retriev purpos applic depend retriev short string popular local search engin yellowpag yahoo local googl map index method need tailor short string purpos number index techniqu relat algorithm propos base length normal similar measur common denomin index length normal measur maintain underli structur presenc increment updat ineffici due data depend precomput weight distinct token string incorpor updat accomplish rebuild index regular time interv paper present framework advoc lazi updat propag kei featur effici increment updat immedi reflect data index strict guarante qualiti subsequ queri answer specif techniqu guarante fals neg limit number fals posit produc implement fulli work prototyp illustr propos idea work practic real dataset queri document experienc unpreced increas content contribut user forum blog social network site microblog servic abund content complement content web site tradit media forum new paper new financi stream plethora inform press cross refer inform textual servic commonli read new item blog report relat content vice versa paper present techniqu autom process cross referenc onlin inform content introduc methodolog extract phrase queri document queri search interfac goal retriev content relat queri document techniqu extract score kei phrase techniqu complement extract phrase inform present extern sourc wikipedia introduc algorithm call relevancerank purpos discuss techniqu detail provid experiment studi util larg number human judg amazon mechan turk servic detail experi demonstr effect effici propos techniqu task autom retriev document relat queri document design queri monitor system optim techniqu reactiv network monitor measur ddriven keyword dqueri expans user gener content fuel explos amount textual data context common user express explicitli numer rate implicitli view opinion product event wealth textual inform necessit develop search data explor paradigm paper propos search model similar spirit facet search enabl progress refin keywordqueri result contrast facet search util domain specif hard extract document attribut refin process driven suggest interest expans origin queri addit search term queri driven domain neutral approach emploi surpris word occurr pattern option numer user rate order identifi meaning top queri expans focu interest subset origin result set propos function support framework computation effici nimbl term storag requir solut ground convex optim principl exploit prune opportun offer natur top formul problem perform benefit offer solut verifi synthet data larg real data set compris blog post improv search social annot data social annot intuit line collabor process element collect resourc url pictur video group descript keyword wide tag group concis accur summari relev resourc content obtain aggreg opinion individu user express form short tag sequenc avail inform rise search paradigm resourc retriev rank base similar keyword queri accompani tag paper present principl effici search resourc rank methodolog util exclus userassign tag sequenc rank base solid probabilist foundat grow understand dynam structur social annot process captur emploi power interpol gram model tag sequenc effici applic propos solut larg data set guarante introduct highli scalabl constrain optim framework emploi train increment maintain gram model experiment valid effici effect solut compar applic approach evalu base larg crawl del icio number hundr thousand user million resourc demonstr applic solut real life larg scale system demonstr interpol gram model tag sequenc result superior rank effect propos optim framework superior term perform obtain rank paramet increment maintain improv perform list intersect list intersect central oper util excess queri process text databas present list intersect algorithm arbitrari number sort unsort list tailor characterist modern hardwar architectur list intersect algorithm present sort list algorithm term dynam probe dynam decid probe order list exploit inform previou probe runtim inform util cach resid microindex algorithm term quantil base deduc advanc good probe order avoid overhead adapt base detect list uniform distribut document identifi unsort list present hashbas algorithm avoid overhead sort detail experiment evalu present base real synthet data exist chip multiprocessor architectur core valid effici efficaci propos algorithm distribut dbase microdata anonym share support ad hoc aggreg analys microdata anonym protect privaci individu varieti privaci model propos microdata anonym model close essenti requir anonym group sensit attribut valu follow distribut support model paper studi problem transform group sensit attribut valu follow target distribut minim data distort specif develop evalu methodolog combin sensit attribut permut gener addit fake sensit attribut valu achiev transform identifi metric relat accuraci aggreg queri answer transform data develop effici anonym algorithm optim accuraci metric varieti data set experiment demonstr effect techniqu anytim measur top dk algorithm exact fuzzi data set valid multi dcolumn schema match type abstract valid multi column schema match essenti success databas integr task difficult databas integr overlap data case practic custom base compani base intuit valu present column relat schema match similar semant type captur distribut valu statist type develop method valid composit schema match techniqu base kei technic idea propos gener measur compar column match schema match base notion inform theoret discrep gener standard geometr discrep basi match present algorithm split string valu column identifi substr match valu column enabl multi column schema match techniqu invalid certif fail valid schema match complement conceptu algorithm contribut experiment studi demonstr effect effici techniqu varieti databas schema data set fast index algorithm set similar select queri abstract data collect inconsist aris due varieti reason desir identifi resolv effici set similar queri commonli data clean match similar data work concentr set similar select queri queri set retriev set collect similar greater threshold set similar measur propos past data clean purpos work concentr weight similar function tf idf introduc variant suit set similar select relat databas context variant special semant properti exploit design effici index structur algorithm answer queri effici present modif exist technolog work set similar select queri introduc algorithm base threshold algorithm exploit semant properti similar measur achiev perform theori practic stretch shrink resiz queri user prefer categor skylin stream data problem skylin comput attract consider research attent categor domain problem complic primarili due partial order natur attribut tupl paper initi studi stream categor skylin identifi limit exist work offlin categor skylin comput realiz techniqu problem maintain skylin categor data stream environ develop lightweight data structur index tupl stream buffer gracefulli adapt tupl attribut partial order domain size complex addition studi domin relat dual space util geometr arrang order index categor skylin effici evalu domin queri lastli experiment studi evalu effici propos techniqu gener target queri databas test tool gener test queri databas explicitli account actual data databas consequ tool guarante suitabl coverag test case commonli requir databas test paper investig problem gener queri satisfi cardin constraint intermedi subexpress execut test databas queri requir test perform databas system oper condit formal analyz problem quantifi difficulti follow analysi descript practic algorithm util sampl space prune techniqu quickli gener test queri desir properti present result experiment evalu approach implement open sourc data manag demonstr util propos ad dhoc aggreg rank list presenc hierarchi adventur blogospher effici sampl inform social network optim join data stream space constrain set select problem hash sampl select estim set similar select queri studi select estim techniqu set similar queri wide varieti similar measur set propos past work concentr class weight similar measur tf idf bm cosin similar variant design select estim base priori construct sampl studi pitfal straightforward applic random sampl argu care sampl construct uniform random sampl yield low accuraci queri sensit realtim sampl expens exact solut cpu cost show build robust sampl priori base exist synops distinct estim prove accuraci techniqu theoret verifi perform experiment algorithm order magnitud faster exact solut small space overhead propag updat spider aggreg queri answer anonym tabl privaci concern microdata releas ad hoc analys privaci goal exist privaci protect approach anonym divers suitabl categor sensit attribut appli directli numer sensit attribut salari result undesir inform leakag propos privaci goal captur privaci protect numer sensit attribut complement desir privaci support ad hoc aggreg analys microdata exist gener base anonym approach answer aggreg queri reason accuraci present gener framework permutationbas anonym support accur answer aggreg queri show group permut base techniqu answer aggreg queri accur gener base approach propos criteria optim permut accur answer aggreg queri develop effici algorithm criterion find skylin top dk bargain solut lightweight onlin framework queri progress indic recent increas interest develop progress indic sql queri paper present lightweight onlin framework problem framework onlin sens refin estim queri progress base feedback receiv queri execut lightweight techniqu design impos minim overhead queri execut sacrif accuraci estim framework estim progress output size relat oper pipelin includ binari multiwai join typic group oper combin thereof describ algorithm effici implement estim present result evalu prototyp implement framework open sourc data manag result demonstr feasibl practic util approach present group linkag author contribut equal manuscript data analysi visual strongli influenc nois nois filter multipl sourc nois microarrai data analysi signal nois ratio rare optim consid report nois analysi million oligonucleotid dataset human featur profil patient muscl biposi recent interact visual tool hierarch cluster explor hce system address effect nois filter resolut arrai correct biolog group unsupervis cluster patient group diagnosi vari probe set interpret method ma rma present call filter cluster linkag method investig result hce hce interact featur enabl quickli impact variabl dendrogram displai show cluster result systemat color mosaic displai provid visual support result show variabl strong effect unsupervis cluster dataset strength biolog variabl maxim nois minim ma present call filter averag group linkag propos gener method interact tool identifi optim signal nois balanc optim combin variabl maxim effect desir biolog variabl data interpret fast identif relat constraint violat benchmark declar approxim select predic pair similar record similar score similar graph blogscop system onlin analysi high volum text stream present blogscop www blogscop net system onlin analysi tempor order stream text appli analysi blogospher system track ten million blog handl hundr thousand updat daili blogscop inform discoveri text analysi system offer set uniqu featur featur includ spatio tempor analysi blog flexibl navig blogospher inform burst keyword correl burst synopsi enhanc rank function improv queri answer relev describ system design featur current version blogscop ad dhoc top dk queri answer data stream top queri retriev highest score tupl data set respect score function defin attribut tupl effici evalu top queri activ research topic instanti problem varieti set studi techniqu develop convent central distribut databas directli applic highli dynam environ line applic data stream recent techniqu support top queri data stream introduc techniqu restrict effici report top answer respect pre oppos ad hoc set queri paper introduc geometr represent top queri problem rais restrict util notion geometr arrang design analyz algorithm increment maintain data set organ arrang represent stream updat introduc queri evalu strategi oper top arrang data structur guarante effici evalu ad hoc queri perform core techniqu augment incorpor tupl prune strategi minim number tupl store manipul result main memori index techniqu support effici increment updat evalu ad hoc top queri experiment studi evalu effici propos techniqu seek stabl cluster blogospher anytim measur top dk algorithm top queri larg multi attribut data set fundament oper inform retriev rank applic paper initi research anytim behavior top algorithm specif top algorithm ta ta sort interest studi progress identif correct result point algorithm execut adopt probabilist approach seek report point oper algorithm confid top result identifi function valuabl asset interest reduc runtim cost top comput present experiment evalu valid techniqu synthet real data set search blogospher abstract happen combin idea social web semant web social web ecosystem particip creat aggreg individu user contribut semant web ecosystem data creat integr structur data sourc applic synthes strength approach creat level rich human particip power structur inform paper propos class applic call collect knowledg system unlock amp quot collect intellig amp quot social web knowledg represent reason techniqu semant web blogscop spatio dtempor analysi blogospher search blogospher abstract happen combin idea social web semant web social web ecosystem particip creat aggreg individu user contribut semant web ecosystem data creat integr structur data sourc applic synthes strength approach creat level rich human particip power structur inform paper propos class applic call collect knowledg system unlock amp quot collect intellig amp quot social web knowledg represent reason techniqu semant web estim select approxim string queri approxim queri string data import due preval data databas convent error string data present vsol estim techniqu estim select approxim string queri vsol estim base invers string make perform select estim independ number string invers string decompos databas string overlap substr length gram associ gram invers string id string gram signatur compress invers string cluster group similar signatur studi techniqu analyt experiment space complex estim depend number neighborhood databas desir estim error time estim select independ number databas string linear wrt length queri string give detail empir perform evalu solut synthet real world dataset show vsol effect larg skew databas short string estim select approxim string queri approxim queri string data import due preval data databas convent error string data present vsol estim techniqu estim select approxim string queri vsol estim base invers string make perform select estim independ number string invers string decompos databas string overlap substr length gram associ gram invers string id string gram signatur compress invers string cluster group similar signatur studi techniqu analyt experiment space complex estim depend number neighborhood databas desir estim error time estim select independ number databas string linear wrt length queri string give detail empir perform evalu solut synthet real world dataset show vsol effect larg skew databas short string index structur match xml twig relat queri processor index structur propos speed evalu xml path express exist xml path indic su er limit focu index structur reli separ index node content simpl path express root leaf path tightli integr relat queri processor unifi framework compar index structur paper present framework defin famili index structur includ exist xml path indic propos index structur famili space time tradeo ectiv evalu xml branch path express twig condit show famili index structur implement access method underli databas system final present experiment evalu show perform tradeo index space match time experiment result show indic achiev order magnitud improv perform evalu twig queri albeit higher space cost previous propos xml path indic tightli integr relat queri processor letter special issu editor publish quarterli distribut tc member scope includ design implement model theori applic databas system technolog letter confer inform new editori revisit machin semant web miss layer human semant web column heterogen measur data qualiti hase hybrid approach select estim conjunct predic syntact rule base approach toweb servic composit reason approxim match queri result join techniqu deploi approxim match predic fundament data clean oper varieti predic util quantifi approxim match oper embed declar data clean framework techniqu return pair tupl relat tag score signifi degre similar tupl pair specif approxim match predic paper problem estim paramet output declar approxim join algorithm plan purpos algorithm highli time consum precis knowledg result size score distribut press concern knowledg aid decis oper promis identifi highli similar tupl kei oper data clean propos solut strategi fulli compli declar framework analyt reason qualiti estim obtain perform strategi present result detail perform evalu strategi propos experiment result valid analyt expect shed addit light qualiti perform estim framework studi offer set simpl fulli declar techniqu problem readili deploi spider declar data clean system rapid identif column heterogen data qualiti concern data manag applic varieti qualiti measur propos accuraci fresh complet captur common sourc data qualiti degrad identifi focu attent measur column heterogen seek quantifi data qualiti problem aris merg data sourc identifi desiderata column heterogen measur intuit satisfi describ techniqu quantifi databas column heterogen base combin cluster entropi soft cluster final present detail experiment result divers data set type demonstr approach robust mechan identifi quantifi databas column heterogen data stream queri process overview paper motiv research issu aris model data process model data form persist relat arriv multipl continu rapid time vari data stream addit review past work relev data stream system current project area paper explor topic stream queri languag requir challeng queri process algorithm issu meta ddata index xpath locat step spider experi report record linkag similar measur algorithm relax join select queri databas user frustrat empti answer queri paper propos framework systemat relax queri involv join select relax queri condit intuit seek minim amount relax yield answer character type answer return relax queri propos lattic base framework order aid queri relax node lattic correspond wai relax queri character properti relax node present algorithm comput answer discuss travers lattic empti queri answer obtain minimum amount queri condit relax implement framework present result perform evalu real synthet data result practic util framework answer top dk queri view consider past work effici comput top object aggreg inform multipl rank list object import instanc problem queri process search engin combin inform post list rank web page object obtain top web page answer user queri studi approach achiev effici top aggreg includ earli termin algorithm ta nra preaggreg input list work rigor treatment combin approach gener ta nra algorithm case preaggreg intersect list addit origin list show version ta nra continu remain instanc optim strong optim notion highlight origin ta nra algorithm index million web page real world search engin queri empir character perform gain offer algorithm show practic benefit intersect list fulli realiz earli termin algorithm similar search match base approach similar search crucial task multimedia retriev data mine exist work model problem nearest neighbor nn problem consid distanc queri object data object fix set featur approach drawback leav partial similar uncov distanc affect dimens high dissimilar overcom drawback propos match problem paper match problem model similar search match queri object data object dimens integ smaller dimension dimens determin dynam make queri object data object return answer set match match queri expect integr xml data sourc approxim join classroom provid copi made distribut pro commerci advantag acm copyright server notic titl public date notic copi permiss acm copi republish post server redistribut list requir prior speci permiss fee approxim stream algorithm histogram construct problem histogram typic approxim data distribut histogram relat synopsi structur success wide varieti popular databas applic includ approxim queri similar search data mine histogram earliest synopsi structur propos continu popular tool typic histogram quick easi estim slight loss accuraci offset fast histogram construct algorithm natur question aris context find fast optim approxim algorithm histogram construct problem paper give linear time factor approxim algorithm gt histogram construct problem algorithm extend data stream show method gener larg number histogram construct problem includ piecewis small degre polynomi approxim data synthet real life data set demonstr approxim histogram ident qualiti scenario offer signific perform benefit letter special issu editor publish quarterli distribut tc member scope includ design implement model theori applic databas system technolog letter confer inform new keyword proxim search xml tree abstract recent work shown benefit keyword proxim search queri xml document addit text document queri keyword shakespear plai xml user interest know keyword cooccur paper focu xml tree defin xml keyword proxim queri return possibl heterogen set minimum connect tree mct match individu keyword queri effici execut keyword proxim queri label tree xml set xml databas preprocess indic xml databas perform detail experiment evalu studi benefit approach show algorithm consider outperform prior algorithm applic approach threshold join algorithm top dk queri distribut sensor network paper present threshold join algorithm tja cient top queri process algorithm distribut sensor network object top queri find highest rank answer user defin similar function evalu queri sensor network environ transfer data extrem expens commun medium tja uniform threshold queri attribut order minim number tupl transfer queri node addition tja resolv queri network central fashion minim consumpt bandwidth delai preliminari experiment result trace driven simul show tja practic cient data stream queri process overview paper motiv research issu aris model data process model data form persist relat arriv multipl continu rapid time vari data stream addit review past work relev data stream system current project area paper explor topic stream queri languag requir challeng queri process algorithm issu adapt process top dk queri xml abil comput top match xml queri gain import due increas number larg xml repositori effici top queri evalu reli score prune irrelev answer earli evalu process context evalu queri plan answer rigid time evalu answer number sequenc oper limit speed score grow adapt queri process permit plan partial match maxim score paper propos architectur adapt algorithm effici comput top match xml queri techniqu evalu exact approxim match approxim defin relax xpath ax order comput score queri answer extend tradit tf idf measur account document structur conduct extens experi varieti benchmark data queri demonstr us adapt approach comput top queri xml monitor dnearest neighbor queri move object locat base applic requir constant monitor nearest neighbor nn queri move object geograph area exist approach problem focus predict queri reli assumpt trajectori object fulli predict queri process time relax index structur match xml twig relat queri processor index structur propos speed evalu xml path express exist xml path indic su er limit focu index structur reli separ index node content simpl path express root leaf path tightli integr relat queri processor unifi framework compar index structur paper present framework defin famili index structur includ exist xml path indic propos index structur famili space time tradeo ectiv evalu xml branch path express twig condit show famili index structur implement access method underli databas system final present experiment evalu show perform tradeo index space match time experiment result show indic achiev order magnitud improv perform evalu twig queri albeit higher space cost previous propos xml path indic tightli integr relat queri processor multipl aggreg data stream monitor aggreg ip traffic data stream compel applic data stream manag system exploratori ip traffic data analysi natur lead pose relat aggreg queri data stream differ choic group attribut paper address problem effici comput multipl aggreg high speed data stream base level lfta hfta dsm architectur inspir gigascop contribut insight scenario addition comput maintain fine granular aggreg queri phantom lfta benefit support share comput contribut investig problem identifi benefici lfta configur phantom user queri formul problem cost optim problem consist optim problem choos phantom alloc space lfta formal show hard determin optim configur propos cost greedi heurist independ problem base detail analys final contribut experiment studi base real ip traffic data synthet data demonstr effect techniqu identifi benefici configur spider flexibl match databas present prototyp system spider develop amp lab research support flexibl string attribut match larg databas discuss design principl spider base describ basic techniqu encompass tool provid descript demo mix meta ddata index system xml present system effici meta data index queri xml document divers inform xml annot xml data wide varieti meta data qualiti secur assess address meta data index problem effici identifi xml element locat step xpath queri satisfi meta data rang constraint system name mix incorpor queri process xpath ax suitabl enhanc metadata featur offer queri answer dynam mainten meta data level xml document approxim join concept techniqu uniti deploi divers approxim match predic object tutori provid comprehens cohes overview kei research result techniqu tool approxim join complement recent data qualiti tutori present broad overview aspect data qualiti don delv detail approxim join technolog permiss copi fee part materi grant provid copi made distribut direct commerci advantag vldb copyright notic titl public date notic copi permiss larg data base endow copi republish requir fee special permiss endow proceed st vldb confer trondheim norwai tutori outlin tutori driven organ formal defin structur content score xml xml repositori queri structur content due structur heterogen xml queri interpret approxim answer return rank score comput answer score xml activ area research oscil pure content score wellknown tf idf take structur account exist propos fulli account structur combin content score queri answer propos xml score method inspir tf idf account structur content queri relax twig score account structur content refer method path score approxim loosen correl queri node reduc amount time requir manipul score top queri process propos effici data structur order speed rank queri process run extens experi valid score method show path score high precis improv score comput time index mix type approxim retriev applic data cleans retriev categor numer attribut base notion approxim match edit distanc numer distanc profound import commonli approxim match predic combin attribut conjunct exist databas techniqu approxim retriev limit applic singl attribut retriev tree variant paper propos methodolog util multidimension index structur problem approxim multi attribut retriev method enabl index collect string numer attribut facilit approxim retriev edit distanc approxim match predic string numer distanc numer attribut approach present base repres set string higher level index structur suitabl compress reason edit distanc queri string compress trie index node feasibl propos evalu techniqu gener compress trie represent fulli index methodolog experiment result show benefit propos compar altern strategi problem effici handl posit predic xml queri process inher order xml document centric data model typic expos posit predic defin xpath navig ax process algorithm axi propos incorpor posit predic receiv attent paper present techniqu leverag power exist state art method effici support posit predic preliminari experiment comparison altern approach reveal perform benefit propos techniqu datacub aggreg approxim queri deviat detect nnh improv perform nearest dneighbor search histogram cient search nearest neighbor nn fundament problem aris larg varieti applic vast practic interest paper propos techniqu call nnh nearest neighbor histogram specif histogram structur improv perform nn search algorithm primari featur propos histogram structur exist conjunct plethora nn search algorithm substanti modifi main idea propos choos small number pivot object space pre calcul distanc nearest neighbor provid complet specif histogram structur show make inform provid ectiv search show construct decid number pivot choos pivot object increment maintain dynam updat util conjunct varieti nn search algorithm improv perform nn search intens experi show nearest neighbor histogram cientli construct maintain conjunct varieti algorithm nn search improv perform dramat ldc enabl search partial distanc hyper ddimension space recent advanc research field multimedia bioinformat brought gener hyper dimension databas hundr thousand dimens hyperdimension databas pose signific problem exist high dimension index techniqu develop index databas commonli hundr dimens support cient queri retriev hyper dimension databas propos methodolog call local digit code ldc support nearest neighbor knn queri hyper dimension databas exist ubiquit indic tree ldc extract simpl bitmap represent call digit code dc point databas prune knn search perform dynam select subset bit dc base subsequ comparison perform expens oper involv comput norm distanc function hyper dimension data avoid extens experi conduct show methodolog er signific perform advantag exist index method real life synthet hyper dimension dataset rout xml queri abstract nowadai increas inform share aris due extens collabor organ organ desir provid data access collabor preserv full control data comprehens privaci user number inform system develop provid effici secur inform share solut propos built atop convent data wareh distribut databas technolog recent inform broker system propos provid privaci preserv inform share loos feder data sourc vulner attribut correl attack queri rout due lack protect rout queri paper investig problem caus attack propos countermeasur limit view queri content intermedi broker show propos content base xpath queri rout scheme level base encrypt commut encrypt effect prevent attribut correl attack origin compromis broker reason overhead kei word inform broker attribut correl attack privaci flexibl string match larg databas practic data clean import process center research interest recent year poor data qualiti result varieti reason includ data entri error multipl convent record databas field signific impact varieti busi issu press technolog enabl flexibl fuzzi match string inform databas cosin similar tf idf establish metric compar text recent propos adapt similar measur flexibl match queri string valu singl attribut relat merg result approxim match oper data clean import process center research interest recent year import end goal effect data clean identifi relat tupl tupl relat queri tupl techniqu propos literatur effici identifi approxim match queri string singl attribut relat addit construct rank order match techniqu associ match score quantifi extent match multipl attribut exist queri tupl issu approxim match oper separ effect creat number rank list relat tupl merg list identifi final rank score return top tupl challeng task paper adapt footrul distanc merg rank list effect deal score studi effici algorithm merg rank produc top tupl declar techniqu approxim match queri string singl attribut relat typic deploi databas introduc describ algorithm problem provid sql specif experiment case studi real applic data realiz propos techniqu commerci data base system highlight benefit propos algorithm attest effect practic approach approxim nn queri stream guarante error fperform bound data stream applic data arriv continu scan queri processor limit memori rel size stream work introduct special issu paper kdd navig index dbase xml multi dqueri process xml path queri form basi complex filter xml data current xml path queri process techniqu divid group navig base algorithm comput result analyz input document tag time contrast index base algorithm advantag precomput number scheme input xml document paper introduc indexbas techniqu index filter answer multipl xml path queri index filter index built document tag avoid process larg portion input document guarante part match analyz index filter compar filter stateof art navig base techniqu show techniqu advantag discuss scenario techniqu superior show xml path queri process techniqu work sax event case pai preprocess input document augment auxiliari inform evalu queri faster present experiment result real synthet xml document valid claim rank join indic order varieti attribut entiti order result effect rank entiti valu attribut domain commonli user correl sourc queri process purpos join oper queri process desir incorpor user prefer specif attribut valu incorpor prefer util score function combin user prefer attribut valu return numer score tupl join result target queri refer top join queri seek identifi tupl join result highest score index dbase approxim xml join xml data integr tool face varieti challeng effici effect oper requir handl varieti inconsist mistak present data set paper studi problem integr xml data sourc index assist join oper notion approxim match structur content xml document join predic show wide deploi index structur tree adopt improv perform oper propos search join algorithm tree adopt index xml document collect propos optim object tree construct make tree suit applic text join data cleans integr rdbm organ data record noisi transcript error incomplet inform lack standard format textual data combin thereof fundament task data clean system match textual attribut refer entiti organ address match effect perform cosin similar metric inform retriev field robust scalabl text join insid rdbm data resid comput exact answer text join expens paper propos approxim samplingbas text join execut strategi robustli execut standard unmodifi rdbm approxim match xml larg amount heterogen inform enterpris data sourc repositori xml data view xml data independ implement paper studi foundat xml data warehous adapt tradit star schema explicit dimens hierarchi xml environ propos notion xml referenti integr handl xml dimens hierarchi queri xml data warehous introduc oper semijoin base approxim match xml data discuss effect evalu correl synchron asynchron data stream panel queri network databas space constrain select problem data warehous pervas comput system keyword proxim search xml databas ding size tree smaller size denot higher associ keyword gener true reason schema design keyword queri yanni vasili permiss copi fee part materi grant provid copi made distribut direct commerci advantag vldb copyright notic titl public date notic copi permiss larg data base endow copi republish requir fee special permiss endow proceed vldb confer berlin germani confer confer confer person sergeabiteboul person chen li person yannispapakonstantin person vasilisvassalo vldb issu year paper titl medmak author icd issu year paper author author sigmod issu year data stream queri process tutori effici approxim optim queri parametr aggreg constraint introduc studi class queri refer opac optim parametr aggreg constraint queri queri aim identifi set databas tupl constitut solut larg class optim problem involv databas tupl constraint object function term aggreg function relat attribut paramet valu identifi constant aggreg constraint develop algorithm preprocess relat construct indic effici provid answer opac queri answer return indic approxim exact provid guarante accuraci indic tune easili meet desir accuraci level provid grace tradeoff answer accuraci index space present result experiment evalu analyz impact paramet accuraci perform techniqu result methodolog effect deploi easili util index structur tree data stream queri process overview paper motiv research issu aris model data process model data form persist relat arriv multipl continu rapid time vari data stream addit review past work relev data stream system current project area paper explor topic stream queri languag requir challeng queri process algorithm issu text join rdbm web data integr integr data produc collect autonom heterogen web servic increasingli import challeng problem due lack global identifi entiti product textual represent databas textual data noisi transcript error incomplet inform lack standard format fundament task data integr match string refer entiti effici bias sampl approxim cluster outlier detect larg data set investig bias sampl densiti data set speed oper gener data mine task cluster outlier detect larg multidimension data set densiti bias sampl probabl point includ sampl depend local densiti data set propos gener techniqu densiti bias sampl factor user requir sampl properti interest tune specif data mine task great flexibl improv accuraci result simpl random sampl describ approach detail analyt evalu show optim approxim cluster outlier detect final present gener substr select estim varieti set relat databas ldap web applic increas quickli accur estim count tupl ldap entri web document match boolean substr queri provid select estim correl occurr substr crucial select estim gener boolean queri studi previous prior work discuss extend appli case onedimension boolean queri exist method case multidimension conjunct queri approxim select explicitli store cross count frequent occur combin substr estim obtain pars queri multidimension substr store cross count appli probabilist formula major problem method number cross count store method increas exponenti number dimens space dimension explos due captur correl dimens limit amount space exist method reliabl give accur estim method gener boolean queri gracefulli present approach select estim gener boolean substr queri focu case conjunct multidimension boolean queri approach explicitli store cross count ddimension substr index structur join primit effici xml queri pattern match xml queri typic pattern select predic multipl element tree structur relationship primit tree structur relationship parent child ancestor descend find occurr structur relationship xml databas core oper xml queri process paper develop famili structur join algorithm task tree merg stack tree tree merg algorithm natur extens tradit merg join recent propos multi predic merg join stack tree algorithm counterpart tradit relat join process present experiment result rang data queri timber nativ xml queri engin built top shore ii commerci relat databas system case structur join algorithm turn vastli superior tradit travers style algorithm structur join algorithm effici tradit join algorithm implement relat databas task final show case tree merg algorithm perform compar stack tree algorithm case consider wors behavior explain analyt result demonstr sort input stack tree algorithm worst case cpu complex linear sum size input output tree merg algorithm guarante approxim data stream queri estim algorithm perform evalu obtain fast good qualiti approxim data distribut problem central interest databas manag varieti popular databas applic includ approxim queri similar search data mine applic domain reli good qualiti approxim histogram base approxim popular method databas theori practic succinctli repres data distribut space effici manner paper place problem histogram construct perspect gener rais requir finit data set data set size case infinit data set data arriv continu form infinit data stream context present singl pass algorithm capabl construct histogram provabl good qualiti present algorithm fix window variant basic histogram construct problem support increment mainten histogram propos algorithm trade accuraci speed grace tradeoff base applic requir case approxim queri infinit data stream present detail experiment evalu compar algorithm applic techniqu real data set demonstr superior propos fast mine massiv tabular data approxim distanc comput stream data manag research direct opportun dlinear dimension reduct techniqu classif visual paper address issu local embed data visual dimens classif advoc basi provid effici map procedur origin dimens data lower intrins dimens depict accur captur user percept similar high dimension data visual purpos exploit low dimension map provid embed develop classif techniqu show experiment classif accuraci compar albeit fewer dimens number classif procedur fast algorithm hierarch rang histogram construct data wareh olap applic typic view data multipl logic dimens product locat natur hierarchi defin dimens olap queri involv hierarch select dimens aggreg measur attribut sale volum accur estim distribut measur attribut hierarch select import varieti scenario includ approxim queri evalu cost base optim queri approxim xml join holist twig join optim xml pattern match inform main inform societi recent develop comput power telecommun constant drop internet access cost data manag store creat condit global diffus web gener research tool analyz inform content depend applic scenario type inform manag search techniqu devis book author deal common type inform plain text discuss part semi structur data xml document deepli discuss part detail analysi approxim match duplic document detect exact approxim semant queri answer multi version document manag person access techniqu offer book guid inform technolog profession user effect effici manag inform knowledg answer increasingli complex dynam multidimension histogram abstract histogram concis flexibl construct summari structur larg data set attract lot attent databas research due util area includ queri optim approxim queri answer basic tool data visual analysi paper present formal studi dynam multidimension histogram structur continu data stream heart propos dynam summari data structur vastli histogram maintain succinct approxim data distribut underli continu stream demand accur histogram deriv dynam data structur propos algorithm extract accur histogram analyz behavior tradeoff propos algorithm provid approxim guarante qualiti estim histogram extract complement analyt result experiment evalu real data set effici approxim scheme data mine task investig bias sampl densiti dataset speed oper gener data mine task cluster outlier detect larg multi dimension dataset densiti bias sampl probabl point includ sampl depend local densiti dataset propos gener techniqu densiti bias sampl factor user requir sampl properti interest tune specif data mine task great flexibl improv accuraci result simpl random sampl describ approach detail analyt evalu show optim approxim cluster outlier detect final present experiment evalu propos method appli densiti bias sampl real synthet data set emploi cluster outlier detect algorithm highlighti count twig match tree describ effici algorithm accur estim number match small node label tree twig larg node label tree summari data structur problem interest queri xml hierarch data provid queri feedback costbas queri optim summari data structur scalabl representsapproxim frequencyinform twiglet small twig data tree twig queri number match estim creat set queri twiglet combin complementari approach set hash estim number match queri twiglet maxim overlap combin queri twiglet estim estim twig queri propos estim algorithm appli approach queri twiglet form variat twiglet decomposit techniqu present extens experiment evalu real xml ddimension substr index effici tunabl similar set retriev model complex data set modern object relat system support set attribut queri capabl paper initi formal studi index techniqu set attribut base similar suitabl defin notion similar set techniqu modern applic recommend collabor filter autom advertis techniqu probabilist approxim natur design principl creat structur make wide data structur techniqu mean eas integr exist infrastructur prefer system effici execut multi dparametr rank queri user optim select object appropri weight import multipl object attribut optim problem oper research appli mathemat everydai life buyer select home weight function number attribut distanc offic price area captur queri definit prefer queri weight function relat attribut deriv score tupl databas system effici produc top result prefer queri evalu weight function tupl relat prefer answer prefer queri effici materi view preprocess entropi base approxim queri explor datacub research devot effici comput relat aggreg specif effici execut datacub oper paper invers problem deriv approxim origin data aggreg motiv problem context specif applic area approxim queri answer data analysi propos framework base notion inform entropi enabl estim origin valu data set aggreg inform show approxim queri data aggreg deriv perform framework describ altern util propos framework enabl identifi valu deviat underli data distribut suitabl data mine purpos final present detail perform studi algorithm real synthet data highlight benefit approach effici propos solut contact author address data dstream histogram histogram wide captur data distribut repres data small number step function dynam program algorithm provid optim construct histogram exist albeit run quadrat time linear space paper provid linear time construct epsilon approxim optim histogram run polylogarithm space result extend context data stream fact gener give epsilon approxim problem data stream requir partit index set interv assumpt requir cost interv monoton inclus larger interv larger cost cost comput approxim small space exhibit nice class problem optim data stream algorithm approxim string join databas free case result return figur queri incomplet su er fals neg contrast claim contrari gij gener string pair omit pair short string string match small edit distanc match meaningless ibm match acm edit distanc absolut fals neg make modif sql queri figur produc correct result fals neg pair short string join pair small string length filter union result result sql queri gij list modifi queri figur experiment result experiment measur number fals neg queri gij figur su er experi thre dgram dbm approxim string process string data ubiquit manag import past year approxim queri import string data due preval typograph error data multipl convent record attribut address commerci databas support approxim string queri directli challeng implement function effici user defin function udf paper develop techniqu build approxim string process capabl top commerci databas exploit facil core techniqu reli gener short substr length call gram process standard method dbm propos techniqu enabl approxim string process method dbm approxim string select join varieti edit distanc function approxim string match predic suitabl edit distanc threshold map vanilla relat express optim convent relat optim reminisc influenti paper dimac summer school tutori frontier data mine space effici bitmap index optim histogram hierarch rang queri tremend interest data wareh olap applic olap applic typic view data multipl logic dimens product locat natur hierarchi de ned select estim boolean queri varieti applic rang optim queri alphanumer attribut provid approxim count document queri term increas quickli reliabl estim number string tupl document match boolean queri boolean queri context consist substr predic compos boolean oper work estim select substr queri gener problem estim select boolean queri substr predic studi approach extract select estim relationship substr predic boolean queri store correl effect multi ddimension index string databas expand scope store pure busi data includ xml document product catalog mail messag directori data increasingli import search databas base wild card string match prefix match common exact match data case match multipl attribut dimens correl dimens tradit multi dimension index structur design fix length numer data mind suitabl match unbound length string data paper describ gener techniqu adapt multi dimension index structur wild card index unbound length string data kei idea carefulli develop map function string ration number repres unbound length string index leaf page fix length offset extern kei store multipl elid identifi repres trend massiv time seri data set sketch index support spatial join high dimension similar join algorithm perform evalu current data repositori includ varieti data type includ audio imag time seri state art techniqu index data queri process reli transform data element point multidimension featur space index queri process place featur space paper studi algorithm find relationship point multidimension featur space specif algorithm multidimension join join convent relat correl multidimension featur space offer valuabl inform data set involv present algorithm paradigm solv multidimension join problem discuss featur limit propos gener size separ spatial join algorithm name multidimension spatial join msj solv multidimension join problem evalu msj specif mine deviant time seri databas high dimension similar join algorithm perform evalu current data repositori includ varieti data type includ audio imag time seri state art techniqu index data queri process reli transform data element point multidimension featur space index queri process place featur space paper studi algorithm find relationship point multidimension featur space specif algorithm multidimension join join convent relat correl multidimension featur space offer valuabl inform data set involv present algorithm paradigm solv multidimension join problem discuss featur limit propos gener size separ spatial join algorithm name multidimension spatial join msj solv multidimension join problem evalu msj specif optim histogram qualiti guarante histogram commonli captur attribut distribut statist queri optim recent histogram consid produc quick approxim answer decis support queri widespread interest histogram motiv problem comput histogram good error metric interest effici algorithm choos bucket boundari minim estim error amount space number bucket convers minim space need upper bound error assumpt find optim bucket boundari computation ineffici previou research focus heurist provabl bound qualiti solut paper present algorithm comput optim bucket boundari time proport squar number distinct data valu size separ spatial join introduc algorithm comput spatial join spatial data set index size separ spatial join impos hierarch decomposit data space contrast previou approach requir replic entiti input data set execut time depend size join data set describ present analyt evalu processor requir compar previous propos algorithm problem show simpl cost estim formula exploit queri optim effici implement softwar present relat system addit introduc dynam spatial bitmap dsb techniqu enabl dynam static exploit bitmap queri process techniqu final present experiment result declust spatial databas multi dcomput architectur present techniqu declust spatial access method share multi comput architectur dg propos softwar architectur tree underli spatial access method leaf level master server leaf node distribut server major contribut work studi optim capac leaf node chunk size stripe unit express respons time rang queri function chunk size show optim implement method network workstat real dataset compar experiment theoret result conclus formula respons time accur maximum rel error typic error vicin illustr wai exploit accur formula examin scenario major practic conclu filter tree manag spatial data rang size granular introduc file organ storag manipul spatial multidimension data execut spatial join oper great effici filter tree inform structur hierarch organ separ spatial entiti size place larger entiti higher level filter tree smaller entiti lower level level index entri entiti order space fill curv hilbert curv algorithm bulk request exploit local index inform minim number transfer disk provid algorithm construct filter tree perform rang queri filter tree perform spatial join pair filter tree full spatial join minimum number index block access need filter tree leaf block read deriv basic properti fil 