evalu combin gener recommend prerequisit optim join map dreduc environ implement map reduc perform oper larg data examin strategi join relat map reduc environ approach begin identifi map kei set attribut identifi reduc process map process send tupl attribut map kei share number bucket valu hash form compon identifi reduc process relat tupl replic limit fashion degre replic depend share map kei attribut miss schema studi problem optim share fix number reduc process algorithm detect fix problem variabl mistakenli includ map kei import special case chain join star join case determin mapkei determin share yield replic method propos superior convent map reduc implement join import case involv larg scale data method win includ analyt queri larg fact tabl join smaller dimens tabl queri involv path graph high degre web social network foundat uncertain ddata integr consider past work studi data integr uncertain data isol develop foundat local view lav data integr sourc integr uncertain motiv distinct set uncertain data integr defin contain uncertain databas set express uncertain sourc view virtual mediat uncertain databas defin consist set uncertain sourc show intract consist check identifi interest special case consist check polynomi final notion answer tradit lav data integr gener uncertain set defin notion correct answer schema design uncertain databas address schema design uncertain databas uncertain data relat natur decomposit kei issu design decomposit reli depend theori primarili function depend studi theori function depend fd uncertain relat defin kind horizont fd vertic fd consist convent fd uncertain relat doesn uncertainti addit standard form decomposit allow ordinari relat fd complex decomposit specif uncertain data show theori fd lossless decomposit uncertain relat present algorithm complex result fundament problem respect fd ordinari uncertain relat test relat instanc satisfi fd find fd satisfi relat instanc infer fd hold result queri uncertain relat fd give sound complet axiomat horizont vertic fd kei special case fd final briefli uncertain data confid valu viewpoint advis student success vol cacm acm org reboot human digit ag view gener effici evalu plan queri gradianc dline acceler learn lowel databas research dassess memori seymour ginsburg survei direct databas system improv effici databas dsystem teach lowel report decad data integr system character classic wrapper mediat architectur base global virtual schema global virtual view gvv propos data sourc store data gvv reconcil integr virtual view underli sourc mediat system propos survei contribut state art advanc focus challeng data integr problem rang syntact semant heterogen present solv heterogen partial manual task requir great amount custom data reconcili write specif reusabl program code lowel report provid guidelin definit public benchmark data integr problem call thalia test lowel databas research assess lowel databas research assess queri websit compact skeleton minim view set lose queri danswer power problem answer queri view studi extens due relev wide varieti data manag applic applic select subset view maintain due limit resourc paper show tradit queri contain good basi decid view select minim view set lose queri answer power formal notion introduc concept contain view set contain view set answer queri answer show contain tradit queri contain relat discuss minim view set retain queri answer power develop idea contain view set respect set queri relationship term queri websit compact skeleton gener effici plan queri view studi problem gener effici equival rewrit view comput answer queri close world assumpt view materi base relat view describ sourc term abstract predic common open world assumpt closedworld model infinit number rewrit comput answer perform queri optim logic plan rewrit queri input gener effici physic plan comput answer goal gener small subset logic plan miss optim physic plan find interest associ support prune abstract associ rule mine heretofor reli condit high support work effici priori algorithm effect rule interest relationship occur frequent number applic data mine identif similar web document cluster collabor filter rule interest compar instanc data case highli correl item possibl causal relationship infrequ item develop famili algorithm solv problem emploi combin random sampl hash techniqu provid analysi algorithm develop conduct experi real synthet data obtain compar perform analysi introduct automata theori languag comput edit databas system implement propos semi automat method translat abstract databas program effici lower level user defin primit present type transform model defin facilit queri translat optim process discuss model framework schema evolut data restructur introduct current databas manag system system base relat model lack express power support complex queri high level languag typic provid limit varieti data structur limit express power high level languag system commonli circumv embed queri languag express convent program languag disadvantag approach forc databas design express conceptu model term semant model addition program languag survei associ drule mine discov associ rule establish problem field data mine exist solut year method propos mine rule sequenti tempor data paper present techniqu base genet program special pattern match hardwar advantag method flexibl adapt abil produc intellig rule consider complex find interest associ support prune abstract associ rule mine heretofor reli condit high support work effici priori algorithm effect rule interest relationship occur frequent number applic data mine identif similar web document cluster collabor filter rule interest compar instanc data case highli correl item possibl causal relationship infrequ item develop famili algorithm solv problem emploi combin random sampl hash techniqu provid analysi algorithm develop conduct experi real synthet data obtain compar perform analysi scalabl techniqu mine causal structur mine associ rule market basket data prove fruit area research measur condit probabl confid correl infer rule form exist item impli exist item rule statist relationship natur relationship presenc presenc convers attribut phenomenon applic know causal relationship extrem enhanc understand effect chang distinguish causal correl difficult problem recent work statist bayesian learn provid avenu attack field goal gener learn complet causal model essenti imposs learn larg scale data mine applic larg number variab inform integr logic view abstract number idea inform integr tool thought construct answer queri view repres capabl inform sourc review formal basi techniqu close relat contain algo rithm conjunct queri datalog program pare approach amp lab amp quot inform manifold amp quot stanford amp quot tsimmi amp quot project term theoret background address inform integr issu review basic idea conjunct queri datalog program contain begin logic rule notat ull amp rule talk edb predic amp quot extension databas amp quot store relat idb predic amp quot intension databas amp quot predic relat construct rule exampl amp quot arc amp quot predic defin graph predic defin structur exist graph mean arc node node case rule amp quot true arc node node arc amp quot repres path length gener atom head left amp quot amp quot sign atom call subgoal side bodi head idb predic subgoal idb edb predic head subgoal assum variabl appear head appear bodi amp quot safeti amp quot requir assur rule left undefin variabl head infer fact head predic assum atom consist predic argument argument variabl constant exclud function symbol argument optim larg join queri mediat system data integr system queri pose mediat translat sequenc queri underli data sourc heterogen environ sourc divers limit queri capabl translat feasibl paper studi problem find feasibl effici queri plan mediat system conjunct queri mediat model sourc capabl attribut bind adorn simpl cost model focus major cost mediat system involv send queri sourc answer back metric develop algorithm sourc queri sequenc base simpl greedi strategi base partit scheme algorithm produc optim plan scenario show linear bound worst case perform miss optim plan algorithm gener optim plan mor mida data dmine project stanford advanc data dmine techniqu abstract comput capabl mediat exist data integr system base mediat architectur emploi avarieti mechan describ queri process capabl sourc system comput capabl mediat base capabl sourc integr paper propos framework captur rich varieti queri process capabl data sourc mediat present algorithm comput set support queri mediat base capabl limit sourc algorithm consider varieti queri process techniqu emploi mediat enhanc set support queri answer queri limit extern queri processor answer queri extern inform sourc content view answer queri rewrit set view present sourc extern inform sourc abil answer limit set queri requir perform oper data set view present sourc infinit albeit encod finit fashion previou work answer queri view consid case set view finit order exploit abil inform sourc answer complex queri problem answer conjunct queri infinit set conjunct view result infinit set conjunct view partit finit number equival class pick view nonempti class suffici determin queri answer view se queri flock gener associ drule mine associ rule mine prove highli success techniqu extract inform larg databas success attribut appropri object fact number queri optim idea priori trick make associ rule mine run faster expect paper trick extend gener context allow effici mine larg databas kind pattern gener idea call queri flock gener test model data mine problem show idea gener purpos mine system gener convent queri optim capabl base mediat tsimmi introduct tsimmi system integr data multipl heterogen sourc user seamless integr view data translat user queri integr view set sourc queri postprocess step comput answer user queri result sourc queri tsimmi mediat architectur accomplish figur user sourc sourc sourc mediat figur tsimmi architectur data integr system garlic inform manifold emploi similar architectur distinguish featur tsimmi semi structur data model call object exchang model oem deal heterogen data sourc emploi sourc wrapper provid uniform oem interfac mediat sigmo comput iceberg queri effici applic comput aggreg function scalabl techniqu mine causal structur mine associ rule market basket data prove fruit area research measur condit probabl confid correl infer rule form exist item impli exist item rule statist relationship natur relationship presenc presenc convers attribut phenomenon applic know causal relationship extrem enhanc understand effect chang distinguish causal correl difficult problem recent work statist bayesian learn provid avenu attack field goal gener learn complet causal model essenti imposs learn larg scale data mine applic larg number variab asilomar report databas research reminisc influenti paper asilomar report databas research databas system index select olap line analyt process olap recent import applic databas system typic olap data present multidimension data cube olap queri complex hour dai run execut directli raw data common method reduc execut time precomput queri summari tabl subcub data cube build index summari tabl commerci olap system todai summari tabl precomput pick select index trial error approach divid space summari tabl index step process perform poorli summari tabl index consum resourc space select effici space paper give algorithm autom select summari tabl index present famili algorithm increas time complex prove strong perform bound algorithm higher complex perform bound increas perform bound diminish show algorithm moder complex perform fairli close optim repres object concis represent semistructur paper introduc repres object uncov inher schema semistructur hierarch data sourc concis descript structur data semistructur data unlik data store typic relat object orient databas fix schema advanc store separ data withth rapid growth world wide web semistructur hierarch data sourc wide casual user lack extern schema inform make brows queri data sourc ineffici imposs worst show repres object make schema discoveri effici facilit gener meaning queri data introduct goal paper present tool repres object facilit queri brows semistructur hierarch inform found web inform integr logic view abstract number idea inform integr tool thought construct answer queri view repres capabl inform sourc review formal basi techniqu close relat contain algo rithm conjunct queri datalog program pare approach amp lab amp quot inform manifold amp quot stanford amp quot tsimmi amp quot project term theoret background address inform integr issu review basic idea conjunct queri datalog program contain begin logic rule notat ull amp rule talk edb predic amp quot extension databas amp quot store relat idb predic amp quot intension databas amp quot predic relat construct rule exampl amp quot arc amp quot predic defin graph predic defin structur exist graph mean arc node node case rule amp quot true arc node node arc amp quot repres path length gener atom head left amp quot amp quot sign atom call subgoal side bodi head idb predic subgoal idb edb predic head subgoal assum variabl appear head appear bodi amp quot safeti amp quot requir assur rule left undefin variabl head infer fact head predic assum atom consist predic argument argument variabl constant exclud function symbol argument dynam itemset count implic rule market basket data problem analyz market basket data present import contribut present algorithm find larg itemset fewer pass data classic algorithm fewer candid itemset method base sampl investig idea item reorder improv low level effici algorithm present gener implic rule normal base anteced consequ implic simpli measur occurr show produc intuit result method final show characterist real data oppos synthet data dramat affect perform system form result introduct area data mine problem deriv associ data recent receiv great deal attent prob tsimmi approach mediat data model languag tsimmi stanford ibm manag multipl inform sourc system integr inform offer data model common queri languag design support combin inform sourc offer tool gener automat compon need build system integr inform paper discuss princip architectur featur rational introduct section introduc core concept mediat tsimmi system implement section discuss compon tsimmi section discuss oem data model tsimmi section introduc mediat specif languag msl section learn gener mediat compon automat final section languag lorel base oem data model tsimmi end user queri languag tsimmi approach mediat data model languag tsimmi stanford ibm manag multipl inform sourc system integr inform offer data model common queri languag design support combin inform sourc offer tool gener automat compon need build system integr inform paper discuss princip architectur featur rational introduct section introduc core concept mediat tsimmi system implement section discuss compon tsimmi section discuss oem data model tsimmi section introduc mediat specif languag msl section learn gener mediat compon automat final section languag lorel base oem data model tsimmi end user queri languag queri semistructur heterogen inform semistructur data absolut schema fix advanc structur irregular incomplet data commonli aris sourc impos rigid structur world wide web data combin heterogen sourc data model queri languag design structur data inappropri environ start lightweight object model adopt tsimmi project stanford paper describ queri languag object repositori design specif semistructur data languag meaning queri result case convent model languag data absent data regular structur similar concept repres type heterogen set present object structur fulli paper motiv kei concept approach describ languag seri mediat languag propos standard type method ffl exchang meta data addit earli maryland meet decid area discuss ffl exchang format data ffl commun protocol includ session control altern approach primari reason want core languag facilit exchang data question adopt exist standard data format asn iso develop precis purpos simpl request languag need request data format problem solut data sourc simpli dump data prescrib format respond request describ data request obtain data problem effici usabl data languag request formul quer databas approach knowledg represent databas theori commun center pod principl databas system confer long term interest logic repres data inform knowledg pick term boil fact atom rule horn claus approach commun slow steadi prefer build carefulli simpl special case gener idea pai attent effici process queri perform oper fact rule power theori develop begin impact applic inform integr engin datalog term datalog coin refer prolog rule function symbol treat logic program unlik prolog convent leastfix point semant rule rule ancestor written data medmak mediat system base declar specif mediat integr heterogen inform sourc present system declar mediat target integr sourc unstructur semi structur data sourc chang schema illustr main featur mediat specif languag msl show facilit integr describ implement system interpret msl specif effici implement data cube materi view deduct databas challeng answer queri limit extern processor integr inform outerjoin full disjunct anand rajaraman jeffrei ullman depart comput scienc stanford univers anand ullman cs stanford abstract motiv piec tidbit inform found web usabl inform structur problem relat comput natur outerjoin relat preserv connect fact comput term full disjunct galindo legaria led question full disjunct comput sequenc natur outerjoin answer involv concept fagin call fl acycl hypergraph prove natur outerjoin sequenc produc full disjunct set relat scheme form connect fl acycl hypergraph motiv imagin construct inform resourc accept queri univers inform gather inform tsimmi databas integr project implement data cube effici decis support applic involv complex queri larg databas respons time small queri optim critic user typic view data multidimension data cube cell data cube view consist aggreg interest total sale valu cell depend valu cell data cube common power queri optim techniqu materi cell comput raw data time commerci system differ approach materi data cube paper investig issu cell view materi expens materi view lattic framework express depend view present greedi algorithm work lattic determin good set view materi greedi algorithm perform small constant factor lore lightweight object repositori semistructur data number inform sourc access electron grow rapidli sourc store export unstructur data addit structur data case unstructur data devoid structur data semistructur data semistructur schema xed advanc data incomplet irregular html le world wide web structur data irregular incomplet addit data integr multipl heterogen inform sourc semistructur store queri semistructur data pose consider di erent problem requir move databas theori databas practic databas research achiev opportun st centuri paper report find queri translat scheme rapid implement wrapper wrapper provid access heterogen inform sourc convert applic queri sourc specif queri command paper present wrapper implement toolkit facilit rapid develop wrapper focu queri translat compon toolkit call convert convert take input queri descript translat languag qdtl descript queri process underli sourc base descript convert decid applic queri directli support translat queri underli system instruct qdtl descript logic support logic equival directli support queri indirectli support comput appli filter automat gener convert result directli support queri introduct wrapper translat pgmw queri semistructur heterogen inform semistructur data absolut schema fix advanc structur irregular incomplet data commonli aris sourc impos rigid structur world wide web data combin heterogen sourc data model queri languag design structur data inappropri environ start lightweight object model adopt tsimmi project stanford paper describ queri languag object repositori design specif semistructur data languag meaning queri result case convent model languag data absent data regular structur similar concept repres type heterogen set present object structur fulli paper motiv kei concept approach describ languag seri version number scheme lexicograph order describ number scheme version altern lexicograph order version hierarchi tree inspect version number easili determin version ancestor determin version sequenc version determin recent common ancestor version upper bound lub sort version number lexicograph result version descend preced ancestor represent nonneg integ delimit lexicograph order match order introduct motiv project collabor design civil engin carri stanford approach project name cedb collabor environ design build lead interest problem version number approa tsimmi approach mediat data model languag tsimmi stanford ibm manag multipl inform sourc system integr inform offer data model common queri languag design support combin inform sourc offer tool gener automat compon need build system integr inform paper discuss princip architectur featur rational introduct section introduc core concept mediat tsimmi system implement section discuss compon tsimmi section discuss oem data model tsimmi section introduc mediat specif languag msl section learn gener mediat compon automat final section languag lorel base oem data model tsimmi end user queri languag answer queri templat bind pattern anand rajaraman yehoshua sagiv jeffrei ullman depart comput scienc stanford univers abstract integr heterogen inform resourc case sourc limit kind queri answer queri ask entir system kind optim problem express queri term limit queri templat sourc answer case conjunct queri show decid nondeterminist polynomi time algorithm queri answer extend result arithmet comparison queri templat motiv data integr system tsimmi papakonstantin garcia widom chawath al translat inform sourc arbitrari type common data model languag sourc sql databas interfac tsimmi inform translat role theori todai survei deduct databas system argument reduct factor identifi properti program spect predic call factor prove detect factor undecid gener show larg class program pro gram obtain appli magic set transfor mation factor respect recurs predic factor properti hold sim ple optim program gener magic set transform result program effici magic set pro gram dramat effici due reduct ariti recurs predi cate show concept factor gener iz previous identifi special case recur sion includ separ recurs left linear recurs special evalua tion algorithm rewrit strategi develop valid constraint partial inform research overview interest problem valid consist integr constraint data modifi constraint check partial inform partial inform includ constraint specif constraint specif modifi data constraint specif modifi data portion exist data method constraint check partial inform effici tradit constraint check method work compil time data access partial inform method enabl constraint check scenario tradit constraint check method fail distribut environ data access explain exist method result queri contain independ appli problem give overview research tsimmi project integr heterogen inform sourc goal tsimmi project develop tool facilit rapid integr heterogen inform sourc includ structur unstructur data paper overview project describ compon extract properti unstructur object translat inform common object model combin inform sourc brows inform manag constraint heterogen site tsimmi joint project stanford ibm almaden research center constraint check partial inform ashish gupta yehoshua sagivi jeffrei ullman jennif widom dept comput scienc stanford univ stanford ca email agupta sagiv ullman widom cs stanford contact phone fax introduct gener problem address tell constraint remain satisfi databas check constraint databas complex answer queri desir minim work check extrem updat guarante creat violat constraint extrem case check effici guarante violat occur expens comput need check constraint test fail constraint constraint queri result ari predic call panic queri produc databas effici complet test databas integr constraint check logic queri optim proff dtree transform frontier databas system research interfac languag theori databas theori gener conjunct queri contain view mainten integr constraint verif abstract comparison deduct object dorient databas system author pretend unbias matter compar deduct object orient approach databas system believ import concept origin object orient approach deduct famili system ultim domin explor reason approach mix fortun featur object orient databas system incorpor deduct system vice versa argu declar import applic declar languag object orient nontrivi contrast approach regard classifi data element close examin prospect classless data motiv prospect databas number distinct class larg systemat number challeng implementin input foutput complex transit closur databas system achiev opportun high dprobabl parallel transit dclosur algorithm input foutput complex transit closur high dprobabl parallel transit closur algorithm select databas research stanford databas system achiev opportun lagunita report nsf invit workshop futur databas system research held palo alto deduct databas achiev futur direct recent year deduct databas focu intens research brought dramat advanc theori system applic salient featur deduct databas capabl support declar rule base style express queri applic databas find applic dispar area knowledg mine databas comput aid design manufactur system paper briefli review kei concept deduct databas newli develop enabl technolog describ current research extend function usabl deduct databas provid synthesi deduct databas procedur object orient approach deduct databas achiev futur direct recent year deduct databas focu intens research brought dramat advanc theori system applic salient featur deduct databas capabl support declar rule base style express queri applic databas find applic dispar area knowledg mine databas comput aid design manufactur system paper briefli review kei concept deduct databas newli develop enabl technolog describ current research extend function usabl deduct databas provid synthesi deduct databas procedur object orient approach bottom dup beat top ddown datalog properti lui moniz pereira anil nerod editor logic program monoton reason proceed intern workshop page lisbon mit press juli min jack minker indefinit databas close world assumpt proceed confer autom deduct york page springer min jack minker foundat deduct databas morgan kaufmann street lo alto ca st edit min jack minker perspect deduct databas journal logic program min jack minker foundat disjunct logic program ew lusk ross overbeek editor proceed north american confer logic program page mit press min jack minker overview nonmonoton reason logic program cs tr comput scienc dept univ maryland univers marylan principl databas knowledg dbase system volum ii bottom dup beat top ddown datalog properti lui moniz pereira anil nerod editor logic program monoton reason proceed intern workshop page lisbon mit press juli min jack minker indefinit databas close world assumpt proceed confer autom deduct york page springer min jack minker foundat deduct databas morgan kaufmann street lo alto ca st edit min jack minker perspect deduct databas journal logic program min jack minker foundat disjunct logic program ew lusk ross overbeek editor proceed north american confer logic program page mit press min jack minker overview nonmonoton reason logic program cs tr comput scienc dept univ maryland univers marylan proof dtree transform theorem applic effici evalu left mult dlinear rule argument reduct factor identifi properti program spect predic call factor prove detect factor undecid gener show larg class program pro gram obtain appli magic set transfor mation factor respect recurs predic factor properti hold sim ple optim program gener magic set transform result program effici magic set pro gram dramat effici due reduct ariti recurs predi cate show concept factor gener iz previous identifi special case recur sion includ separ recurs left linear recurs special evalua tion algorithm rewrit strategi develop converg queri evalu optim synchron hypercub principl databas knowledg dbase system volum modern databas manag system essenti solv problem access manag larg volum relat data singl platform cluster tightli coupl platform problem remain databas work fundament problem rais semant heterogen fact data du plicat multipl databas repres differ underli databas schema tutori describ fundament problem rais semant heterogen survei theoret framework provid solut tutori consid topic repr sent architectur support databas interoper notion compar inform capac databas schema provid support read integr view data includ virtual materi approach provid support read write integr view data includ issu workflow heterogen databas research tool access effect meta data identifi relationship schema databas complex order subgoal parallel complex logic queri program effici test top ddown termin logic rule optim synchron hypercub databas theori past futur algorithm compil regular express pla yawn window nail commun dtime tradeoff compil princil techniqu tool parallel complex logic queri program design overview nail system paper adopt view model theoret circumscript cf etherington horti literatur restrict interest herbrand pre interpret notwithstand comput complex circumscript algorithm develop comput circumscript true algorithm poor worst case complex work practic experiment result prototyp implement linear program base comput monoton databas effici practic prolog precis detail experi found similar vein precis effici techniqu comput circumscript judg implement prototyp system carri detail experiment implement develop prototyp implement comput circumscript detail implement experiment result report precis effici optim report part ii seri paper preliminari experiment result algorithm analog algorithm comput parallel priorit circumscript time spent iter loop compact method yield signific save time concept call local split databas manag chunk base depend analysi yield signific save time requir compil experiment result preliminari stage expect month reliabl statist obtain magic set strang wai implement logic program sld resolut negat finit failur sldnf reflect procedur interpret predic calculu program languag form comput basi prolog system advantag stack base memori manag sldnf queri evalu reason termin due infinit posit recurs termin due infinit recurs negat repeatedli evalu liter rule bodi lead unaccept perform address problem fir goal orient queri evalu gener logic program present tabl evalu delai slg resolut converg sidewai queri evalu approach process queri logic dbase queri languag updat logic databas suggest approach databas updat databas treat collect theori investig issu simultan multipl updat oper equival databas updat oper logic databas system recent work queri optim address issu place expens predic queri plan paper explor predic placement option consid montag dbm present famili algorithm form success complex effect optim solut analysi perform measur montag sql queri classifi queri highlight simplest solut optim class correctli demonstr limit previous publish algorithm discuss challeng feasibl implement algorithm commerci grade system introduct relat databas manag system begun user defin data type oper util ad hoc queri simultan objectori dbmss begun offer ad hoc queri facil allow declar access object method previous access handcod imper appl approach process queri logic dbase queri languag commun dtime tradeoff implement logic queri languag databas abstract implement logic queri languag databas thought supercomput organ commun dtime tradeoff equival logic databas complementari independ map databas correct fault write donc memori foundat univers relat model system fu databas system base univers relat assumpt flux sort supercomput organ ai applic connect acycl hypergraph place ball independ uniformli random bin fullest bin hold theta log log log ball high probabl recent azar al analyz randomli choos bin ball sequenti place ball full chosen bin show fullest bin log log log theta ball high probabl explor extens result parallel distribut set result focu tradeoff amount commun final load round commun provid lower bound maximum load omega gamma log log log wide class strategi result extend case number round allow grow demonstr parallel sequenti strategi present azar al achiev load constant factor lower bound data structur algorithm univers relat interfac databas system reveng jd semant updat databas suggest methodolog updat data abil databas user lo modifi con base integr constrain amp rule deriv inex tent databas call updal oper fun plicit inform problem dament databas manag system updat arbitrari theori insert inu delet user deal entir conceptu databas fragment relat applic domain todai databas size grow rapidli pose great demand technolog data fragment facilit techniqu distribut parallel main memori comput meet demand fragment improv cient process queri type top databas design queri optim requir good notion cost result fragment mathemat deriv select model facilit paramet comput base fragment infrequ updat model forget data distribut result fast good select estim show experiment verif zipfian distribut ir databas keyword select fragment zipf inform retriev databas introduct cient ectiv process larg amount data crucial import comput applic notion inform transfer vlsi circuit maxim object semant univers relat databas corrigendum theori join relat databas kent consequ assum univers relat univers algorithm sequenti data compress present perform investig respect nonprobabilist model constrain sourc compress ratio achiev propos univers code uniformli approach lower bound compress ratio attain block variabl code variabl block code design match complet sourc tool templat depend principl databas system edit regular express compil develop gigascop stream databas network applic includ traffic analysi intrus detect router configur analysi network research network monitor perform monitor debug gigascop undergo instal site amp network includ oc router detail monitor paper describ motiv constraint develop gigascop gigascop architectur queri languag perform issu conclud discuss stream databas research problem found applic strike back connect acycl hypergraph place ball independ uniformli random bin fullest bin hold theta log log log ball high probabl recent azar al analyz randomli choos bin ball sequenti place ball full chosen bin show fullest bin log log log theta ball high probabl explor extens result parallel distribut set result focu tradeoff amount commun final load round commun provid lower bound maximum load omega gamma log log log wide class strategi result extend case number round allow grow demonstr parallel sequenti strategi present azar al achiev load constant factor lower bound simplifi univers relat assumpt properti problem univers relat assumpt inabl method obtain databas scheme design gener case real world constraint set depend includ embed multivalu depend propos simpler method describ real world constraint function depend singl join depend relationship method defin real world classic method expos character term hypergrapb multivalu depend consequ join depend character term hypergraph join depend equival set multivalu depend queri process univers relat system templat depend larg class depend relat databas complet axiomat analysi design relat schemata databas system phd thesi ucla juli fagin fagin normal form relat databas base domain kei acm transact databas system septemb gadia gadia homogen relat model queri languag tempor databas acm transact databas system decemb gadia gadia seamless gener extens sql queri tempor data technic report tr comput scienc depart iowa state univers jensen al jensen mark roussopoulo increment implement model relat databas transact time ieee transact knowledg data engin decemb jensen amp snodgrass jensen snodgrass propos data model tempor struc compil regular express integr circuit theori function templat depend view direct relat databas theori optim wire rectangl properti acycl databas scheme connect acycl hypergraph place ball independ uniformli random bin fullest bin hold theta log log log ball high probabl recent azar al analyz randomli choos bin ball sequenti place ball full chosen bin show fullest bin log log log theta ball high probabl explor extens result parallel distribut set result focu tradeoff amount commun final load round commun provid lower bound maximum load omega gamma log log log wide class strategi result extend case number round allow grow demonstr parallel sequenti strategi present azar al achiev load constant factor lower bound fragment relat hack equival relat databas scheme modern databas manag system essenti solv problem access manag larg volum relat data singl platform cluster tightli coupl platform problem remain databas work fundament problem rais semant heterogen fact data du plicat multipl databas repres differ underli databas schema tutori describ fundament problem rais semant heterogen survei theoret framework provid solut tutori consid topic repr sent architectur support databas interoper notion compar inform capac databas schema provid support read integr view data includ virtual materi approach provid support read write integr view data includ issu workflow heterogen databas research tool access effect meta data identifi relationship schema databas infer tree lowest common ancestor applic optim relat express deadlock dfree packet switch network abstract deadlock system failur observ exist packet switch network psn problem relat design deadlock free psn investig polynomi time algorithm problem shown np complet np hard polynomi time algorithm found kei word packet switch network deadlock complex np complet flow control rout comput network commun network basic definit packet switch network psn direct graph vertic repres processor edges repres commun link assum messag call packet pass processor vertex vi constant bi number buffer vertex buffer hold packet packet acycl rout vq path vertex sourc vq destin vertex packet assum afixedroutingprocedur kl apacket rout determin sourc node assum rout packet includ part messag packet practic packet principl databas system st edit compil regular express integr circuit extend abstract interact function depend templat depend abstract paper discuss algorithm topic select queri design queri databas metadata web inform resourc metadata databas topic relationship call metalink topic topic databas import score topic select oper tselect select time topic satisfi select formula output import score threshold top select formula expens predic form user defin function minim number expens predic evalu probe tselect algorithm introduc evalu heurist due time constraint tselect algorithm termin locat output tupl order maxim number output tupl found introduc evalu heurist locat tupl evalu time complet axiomat larg class depend relat databas system fu databas system base univers relat assumpt system fu databas system base univers relat assumpt univers instanc scheme automat infer variabl type adequaci decomposit relat databas introduct automata theori languag comput adequaci decomposit relat databas model commun protocol automata univers data retriev languag structur analys softwar system natur formal relat queri detect design pattern pattern problemat design code clone dead code di erenc built design architectur paper describ crocopat applic independ tool relat program cienci express languag crocopat enabl practic import analys real world softwar system graph analysi tool analys involv transit closur detect pattern graph languag easi base order predic logic tool easi integr softwar system small command line tool simpl text format input output relat equival relat databas scheme modern databas manag system essenti solv problem access manag larg volum relat data singl platform cluster tightli coupl platform problem remain databas work fundament problem rais semant heterogen fact data du plicat multipl databas repres differ underli databas schema tutori describ fundament problem rais semant heterogen survei theoret framework provid solut tutori consid topic repr sent architectur support databas interoper notion compar inform capac databas schema provid support read integr view data includ virtual materi approach provid support read write integr view data includ issu workflow heterogen databas research tool access effect meta data identifi relationship schema databas deadlock dfree packet switch network abstract deadlock system failur observ exist packet switch network psn problem relat design deadlock free psn investig polynomi time algorithm problem shown np complet np hard polynomi time algorithm found kei word packet switch network deadlock complex np complet flow control rout comput network commun network basic definit packet switch network psn direct graph vertic repres processor edges repres commun link assum messag call packet pass processor vertex vi constant bi number buffer vertex buffer hold packet packet acycl rout vq path vertex sourc vq destin vertex packet assum afixedroutingprocedur kl apacket rout determin sourc node assum rout packet includ part messag packet practic packet optim partial dmatch retriev field independ theori join relat databas effici optim class relat express equival relat express present structur logist regress extens logist regress model relat data integr approach build regress model data store relat databas potenti predictor boolean real valu gener structur search space queri databas test statist inform criteria inclus logist regress statist relat represent model noisi domain complex structur link predict task high interest characterist domain scientif citat social network hypertext underli data extrem noisi featur predict readili flat file format propos applic structur logist regress build link predict model present experiment result task predict citat made scientif literatur relat data cites search engin data includ citat graph authorship public venu paper word content crossfertil dbm research disciplin comput scienc panel discuss gener scheme automat infer variabl type effici optim class relat express abstract theori join relat data base extend abstract code gener machin multiregist oper entiti resolut receiv consider attent recent year refer underli entiti goal predict refer correspond entiti show extend latent dirichlet alloc model task propos probabilist model collect entiti resolut relat domain refer connect approach differ recent propos entiti resolut approach gener make pair wise decis captur relat entiti hidden group variabl propos sampl algorithm collect entiti resolut unsupervis take entiti relat account addition assum domain entiti show infer number entiti data demonstr util practic relat entiti resolut approach author resolut real world bibliograph dataset addit present preliminari result character condit relat inform model statist databas secur monoton data flow analysi framework oper spars relat code gener express common subexpress find depth flow graph evalu relat express dens spars argument economi descript parser dpda pda induct variabl high level languag code gener express common subexpress find depth flow graph protect oper system exist mandatori access control system oper system difficult identifi principl design usabl access control system introduc usabl mandatori integr protect umip model add usabl mandatori access control oper system umip model design preserv system integr face network base attack usabl goal umip twofold configur umip system difficult instal configur oper system exist applic common usag practic umip umip featur achiev goal introduc concept express partial trust program leverag inform exist discretionari access control mechan deriv file label mandatori integr protect discuss implement umip model linux linux secur modul framework show simpl configur low overhead effect defend number network base attack bound complex longest common subsequ problem survei theori practic sequenc comparison made focus diff unix file differ util sequenc comparison sequenc comparison deep fascin subject comput scienc theoret practic opinion theoret practic aspect problem understood feel masteri true challeng comput scienc central problem state easili find algorithm effici practic comput longest common subsequ lc short sequenc usual subsequ sequenc sequenc obtain delet necessarili contigu term en pri en pai longest common subsequ sequenc comparison theori practic part work author visit universit de rouen visit partial support global data flow analysi iter algorithm abstract kildal develop data propag algorithm code optim gener lattic theoret framework directmn hecht ullman gave strong upper bound number iter requir propag algorithm data repres bit vector depth order flow graph present paper combin idea paper condit whmh bound hecht ullman appli depth veremn klldall gener data propag algorithm shown condit suffici function block flow graph arbitrari lattic element lattic vf fg gt shown instanc techniqu kildal found meet condit node list reduc flow graph find lowest common ancestor tree paper crew pram algorithm problem find lowest common ancestor forest insert leav root delet leav forest maximum vertic algorithm take log min log time space processor process sequenc oper present round lowest common ancestor queri worst case constant time singl processor processor algorithm match bound achiev sequenti algorithm algorithm simpler smaller constant time space complex introduct find lowest common ancestor tree frequent occur problem literatur found applic divers problem comput domin reduc flow graph detect neg cycl spars graph planar test comput weight matc economi descript parser dpda pda evalu relat express dens spars argument applic lattic algebra loop optim complex lr test protect oper system exist mandatori access control system oper system difficult identifi principl design usabl access control system introduc usabl mandatori integr protect umip model add usabl mandatori access control oper system umip model design preserv system integr face network base attack usabl goal umip twofold configur umip system difficult instal configur oper system exist applic common usag practic umip umip featur achiev goal introduc concept express partial trust program leverag inform exist discretionari access control mechan deriv file label mandatori integr protect discuss implement umip model linux linux secur modul framework show simpl configur low overhead effect defend number network base attack node list reduc flow graph complex lr test determinist pars ambigu grammar entiti resolut receiv consider attent recent year refer underli entiti goal predict refer correspond entiti show extend latent dirichlet alloc model task propos probabilist model collect entiti resolut relat domain refer connect approach differ recent propos entiti resolut approach gener make pair wise decis captur relat entiti hidden group variabl propos sampl algorithm collect entiti resolut unsupervis take entiti relat account addition assum domain entiti show infer number entiti data demonstr util practic relat entiti resolut approach author resolut real world bibliograph dataset addit present preliminari result character condit relat inform np dcomplet schedul problem paper present develop simul genet algorithm ga base methodolog appli optim tune fuzzi dispatch system fleet autom guid vehicl flexibl manufactur environ dispatch rule transform continu adapt procedur capit onlin inform shop floor time entir problem simul matlab simulink simul result obtain show ga effici effect tool achiev optim perform np complet schedul problem simpl algorithm global data flow analysi problem evalu polynomi fix set point design analysi comput algorithm bound complex longest common subsequ problem extend abstract oper spars relat effici algorithm grammar problem extend abstract character reduc flow graph worst dcase perform bound simpl ddimension pack algorithm abstract abstract problem model practic problem comput scienc oper research list real number place element ofl minimum number amp quot bin amp quot bin number sum exce motiv likelihood excess amount comput requir algorithm determin optim placement examin perform number simpl algorithm obtain amp quot good amp quot placement fit algarithm place number success bin inwhich fit fitalgorithm placeseachnumb success themostnearli full bin fit show fit fit algorithm bin outlin proof decreas order algorithm thanl bin exampl show upper bound essenti similar result obtain list number larger lt introduct recent result complex theori combinatori optim problemsmai effect imposs solv determinist pars ambigu grammar entiti resolut receiv consider attent recent year refer underli entiti goal predict refer correspond entiti show extend latent dirichlet alloc model task propos probabilist model collect entiti resolut relat domain refer connect approach differ recent propos entiti resolut approach gener make pair wise decis captur relat entiti hidden group variabl propos sampl algorithm collect entiti resolut unsupervis take entiti relat account addition assum domain entiti show infer number entiti data demonstr util practic relat entiti resolut approach author resolut real world bibliograph dataset addit present preliminari result character condit relat inform analysi simpl algorithm global flow problem polynomi complet schedul problem find lowest common ancestor tree paper crew pram algorithm problem find lowest common ancestor forest insert leav root delet leav forest maximum vertic algorithm take log min log time space processor process sequenc oper present round lowest common ancestor queri worst case constant time singl processor processor algorithm match bound achiev sequenti algorithm algorithm simpler smaller constant time space complex introduct find lowest common ancestor tree frequent occur problem literatur found applic divers problem comput domin reduc flow graph detect neg cycl spars graph planar test comput weight matc fast algorithm elimin common subexpress pars algorithm backtrack paper propos effici linguist process strategi speech recognit understand depend structur grammar strategi includ parsingand phrase predict algorithm speech process phrase recognit base phonem recognit parser extract sentenc likelihood take account phonet likelihood phrase candid linguist likelihood semant inter phrase depend relationship fast pars algorithm breadth search propos predictor pre select pt as candid transit rule combin depend structur reduc amount phonet process propos linguist processor test speech recognit experi experiment result show greatli increas accuraci speech recognit breadth pars algorithm predictor increas process speed error detect preced parser techniqu speed lr parser set merg algorithm paper present implement comput model lexic develop case action verb simul agent train inform give label agent action hand motion system learn label carri similar action computation system emploi form activ represent explicitli intend neural plausibl learn methodolog version bayesian model merg omohundro verb learn model broader context project embodi natur languag acquisit fast algorithm elimin common subexpress worst dcase analysi memori alloc algorithm flow graph reduc abstract common iter recurs dsp applic repres synchron data flow graph sdfg great deal research attempt optim applic retim proven effect transform singl rate data flow graph equival dfg smaller clock period retim attempt reduc execut time synchron dfg explor paper develop basic definit result express studi sdfg review problem face attempt retim sdfg order minim clock period present algorithm final demonstr effect method exampl techniqu speed lr parser weak mix strategi preced pars note effici hash function equival program structur variabl optim lr parser sibl jame jr jacquelin josephin educ support persev professor hank dietz advisor encourag toler past year master doctor professor russel quong interest research help develop advanc languag pars idea antlr parser gener acknowledg professor dave meyer jose fort interest electr engin encourag attend graduat school professor leah jamieson deserv take chanc support continu studi purdu professor matt keef paul woodward univers minnesota provid fellowship finish studi paul woodward give work ack acknowledg kevin edgar user purdu compil construct tool set codevelop cohen optim straight line program transit reduct direct graph modern enterpris increasingli workflow paradigm prescrib busi process perform process typic model annot activ graph present approach system construct process model log past unstructur execut process graph produc conform depend past execut present log provid model captur previou execut process techniqu easier introduct workflow system evalu evolut exist process model present result appli algorithm synthet data set process log obtain ibm flowmark instal flow graph reduc abstract common iter recurs dsp applic repres synchron data flow graph sdfg great deal research attempt optim applic retim proven effect transform singl rate data flow graph equival dfg smaller clock period retim attempt reduc execut time synchron dfg explor paper develop basic definit result express studi sdfg review problem face attempt retim sdfg order minim clock period present algorithm final demonstr effect method exampl care feed lr grammar translat context dfree grammar principl optim page replac character extens pushdown translat pars algorithm backtrack paper propos effici linguist process strategi speech recognit understand depend structur grammar strategi includ parsingand phrase predict algorithm speech process phrase recognit base phonem recognit parser extract sentenc likelihood take account phonet likelihood phrase candid linguist likelihood semant inter phrase depend relationship fast pars algorithm breadth search propos predictor pre select pt as candid transit rule combin depend structur reduc amount phonet process propos linguist processor test speech recognit experi experiment result show greatli increas accuraci speech recognit breadth pars algorithm predictor increas process speed equival program structur variabl transform straight line program dpreliminari version gener optim code arithmet express comput program portabl extent move comput environ effort rewrit limit program perfectli portabl move chang whatsoev recent languag extens made easier write portabl program tool develop aid detect nonport construct tool program move pdp develop machin unix oper system softwar transport interdata sourc languag represent code involv ident environ comput power pushdown automata character dwai determinist class languag character dwai determinist class languag translat context free grammar result tape dbound ture machin halt stack automata syntax direct translat pushdown assembl commun work rapidli grow field statist tree structur model properti syntax direct translat concept macro tree transduc formal model studi properti syntax direct translat function languag base primit recurs paper output languag produc nondelet macro tree transduc introduc prove pump lemma appli pump lemma gain result produc nondelet macro tree transduc comput set monad tree doubl exponenti height output hierarchi produc nondelet macro tree transduc respect number function gener theori translat order defin semant diagram languag techniqu develop establish approach denot oper algebra semant program languag due multi dimension natur diagram oppos linear structur program approach base graph tree term graph transform provid underli technolog paper substanti claim review import approach semant discuss applic diagram languag keyword diagram languag semant graph transform introduct import mean commun languag languag natur human human commun artifici human comput comput comput commun gener case speak write read symbol la automaton analog syntax direct translat schemata univers state machin character theorem decomposit scheme set accept dwai stack automata context sensit time tape complex pushdown automaton languag decid undecid question automata relat time tape complex determinist stack automata quotient oper theori languag studi random version comput model introduc restrict random access extern memori intern memori space essenti model view power version data stream model put cost sequenti scan extern memori model data stream addit extern memori model unlik stream model admit larg extern memori devic read written parallel obtain tight lower bound decis problem set equal multiset equal checksort precis show random side error bound mont carlo algorithm problem perform logn random access extern memori devic provid intern memori size logn denot size input data lower bound set equal problem infer lower bound worst case data complex queri evalu languag xqueri xpath relat algebra stream data precis show exist queri xqueri xpath relat algebra random la vega algorithm evalu queri perform logn random access extern memori devic provid intern memori size logn approach unifi theori automata result dwai stack automata noneras stack automata 