page support effici docq prototyp system support data dorient content queri data dorient content queri system search data text web aid ad dhoc intent detect engin queri log deep dweb search paper present survei major approach search interfac understand deep web consist data exist web inaccess text search engin tradit access data manual fill html form search interfac scalabl grow size deep web automat access data requir automat understand search interfac easi human perceiv interfac machin process interfac challeng decad work address automat interfac understand problem emploi varieti understand strategi paper present survei integr web queri result holist schema match probabilist top dk rank daggreg queri trustworthi keyword search complianc storag advanc web network technolog entiti search engin agil deffort inform integr web collabor wrap turbo framework web data extract dewex explor facil enabl deep web integr top dk queri process uncertain databas top process uncertain databas semant computation tradit top process interplai score uncertainti make tradit techniqu inapplic introduc probabilist formul top queri formul base marriag tradit top semant world semant light formul construct framework encapsul state space model effici queri process techniqu tackl challeng uncertain data set prove techniqu optim term number access tupl materi search state experi show effici techniqu data distribut order magnitud improv na ve materi world progress select merg comput top dk ad dhoc rank function famili threshold algorithm ta wide studi effici comput top queri ta sort merg framework assum data list pre sort rank function monoton databas applic attribut valu index treestructur indic tree tree rank function necessarili monoton answer top queri ad hoc rank function paper studi index merg paradigm perform progress search space joint state compos multipl index node address challeng effici queri process minim search complex present doubleheap algorithm support progress state search progress state gener avoid unnecessari disk access character type empti state contribut final result propos materi model join signatur prune empti state perform studi show propos method achiev order magnitud speed baselin solut urank formul effici evalu top dk queri uncertain databas top process uncertain databas semant computation tradit top process interplai queri score data uncertainti make tradit techniqu inapplic introduc urank system process probabilist formul top queri uncertain databas formul base marriag tradit top semant world semant urank encapsul process framework leverag exist queri process capabl implement effici search strategi integr rank score rank probabl obtain meaning answer top queri support entiti search larg dscale prototyp search engin support rank cluster gener order dby group dby boolean semant sql queri adequ captur fuzzi prefer soft criteria requir tradit data retriev applic solv problem add flavor inform retriev databas queri allow fuzzi queri condit flexibl support group rank queri result dbm engin rank support major commerci dbmss nativ support flexibl group limit group paper propos gener group enabl flexibl group cluster specif queri result cluster data mine applic focu support effici cluster boolean result gener queri time propos integr rank cluster boolean condit form type clusterrank queri structur data retriev integr nontrivi term semant queri process investig semant type queri process queri straightforward approach simpli glue techniqu develop rank cluster approach costli rank cluster treat block post process task boolean queri result exist techniqu propos summari base evalu method util bitmap index seamlessli integr boolean condit cluster rank experiment studi show approach significantli outperform straightforward maintain high cluster qualiti entityrank search entiti directli holist web evolv data rich repositori standard page view current search engin increasingli inadequ wide rang queri task search data entiti phone number paper pdf date todai engin indirectli page entiti page current engin find page individu search directli holist find inform finer granular studi problem entiti search signific departur tradit document retriev focu core challeng rank entiti distil underli conceptu model impress model develop probabilist rank framework entityrank seamlessli integr local global inform rank evalu onlin prototyp tb web corpu show entityrank perform effect context dawar wrap synchron data extract deep web present press integr larg number dynam evolv data sourc automat accur build integr system observ problem sequenti task integr wrapper extract task peer sourc facilit subsequ match task parallel sourc wrapper leverag peer wrapper domain rule enhanc extract accuraci issu seemingli unrel boil lack context awar current automat wrapper induct approach gener wrapper sourc time isol inher lack awar peer sourc domain knowledg context integr propos concept context awar wrapper amen match leverag peer wrapper prior domain knowledg context awar inspir synchron framework construct wrapper consist collabor mutual context draw insight turbo code develop turbo syncer interconnect extract match achiev context awar wrap experi show turbo syncer hand enhanc extract consist increas match accuraci measur hand incorpor peer wrapper domain knowledg seamlessli reduc extract error optim top dk queri middlewar access unifi cost dbase approach paper studi optim top queri middlewar assort algorithm propos gener applic wide rang scenario exist algorithm lack gener support wide rang access scenario systemat adapt account runtim specif fulfil critic lack aim take cost base optim approach runtim search space algorithm cost base optim gener wide rang access scenario adapt specif access cost runtim optim grant relat queri earli lack rank queri paper identifi address barrier realiz unifi framework barrier defin comprehens space encompass possibl optim algorithm search barrier conflict goal space focus enabl effici search sql queri explicitli compos relat oper space definit consist schedul oper queri plan contrast top queri notion logic task unit schedul defin logic task top queri build block identifi comprehens focus space top queri develop effici search scheme space identifi optim algorithm studi framework unifi outperform exist algorithm specif design scenario access deep web web rapidli deepen myriad searchabl databas onlin enabl effect access deep web build metaqueri explor integr databas web metaqueri tackl integr larg scale sourc prolifer onlin dynam natur queri access sourc integr approach hing insight challeng larg scale opportun observ desir semant connect surfac present characterist hidden regular sourc gener recent work paper propos approach shallow integr holist sourc discov desir semant exploit hidden regular shallow clue sourc holist probe minim schedul optim support top dk queri expens predic abstract paper address problem evalu rank top queri expens predic major dbmss support expens user defin predic boolean queri support rank queri import rank queri model user specif concept prefer relev similar call dynam user defin function middlewar system incorpor extern predic integr autonom sourc typic access object queri rank queri accompani boolean rank condit turn predic expens index structur predic built base tabl longer effect retriev filter object order fourth fuzzi join inher expens essenti user defin oper dynam associ multipl relat predic dynam defin extern access reli index mechan provid time sort output requir object probe evalu enabl probe minim develop problem cost base optim search potenti probe schedul decoupl probe schedul object predic schedul problem develop analyt object schedul optim dynam predic schedul optim combin form cost effect probe schedul index term databas queri process distribut inform system databas system enabl soft queri data retriev boolean rank queri databas dconstrain optim support ad dhoc rank aggreg paper present principl framework effici process ad hoc top rank aggreg queri provid group highest aggreg result essenti support queri lack current system process queri na ve materi group sort scheme prohibit ineffici framework base fundament principl upper bound principl dictat requir earli prune group rank tupl rank principl dictat group order tupl order requir guid queri processor provabl optim tupl schedul aggreg queri process propos execut framework appli principl requir address challeng realiz framework implement queri oper enabl effici group awar rankawar queri plan experiment studi valid framework demonstr order magnitud perform improv queri plan compar tradit plan access web search integr wit rapid growth world wide web web broaden deepen survei estim total million page web time nowadai billion web page report recent index yahoo denot time increas year surfac web link billion static html page equal signific amount inform hidden deep web queri form searchabl databas juli survei estim search site billion content page deep web recent studi april estim deep web sourc million queri form reflect fast time increas year virtual unlimit amount unstructur structur automat complex schema match web queri interfac correl mine approach enabl inform integr schema match critic step discov semant correspond attribut heterogen sourc complex match common complex search space exist techniqu focu simpl match tackl challeng articl take conceptu approach view schema match correl mine task match web queri interfac integr myriad databas internet deep web queri interfac gener form complex match attribut group author correspond book domain observ occurr pattern queri interfac reveal complex semant relationship group attribut tend present queri interfac posit correl contrast synonym attribut neg correl rare occur insight enabl discov complex match correl mine approach develop dcm framework consist data preprocess dual mine posit neg correl final match construct evalu dcm framework manual extract interfac result show good accuraci discov complex match autom larg scale integr build metaqueri databas web web rapidli deepen myriad searchabl databas onlin data hidden queri interfac larg scale integr deep web build metaqueri system explor find integr queri databas web interim report paper propos goal metaqueri web scale integr dynam ad hoc natur larg scale integr mandat dynam sourc discoveri thefli queri translat present system architectur underli technolog kei subsystem ongo implement build metaqueri extract match web queri interfac introduct recent wit rapid growth preval databas web recent studi april estim onlin databas deep web myriad databas provid dynam queri base data access queri interfac static url link essenti integr queri interfac integr deep web goal metaqueri project http metaqueri cs uiuc aim open deep web user build system user explor integr deep web sourc start focu integr deep web sourc domain import integr task typic scenario includ purchas book lowest price book sourc flight ticket tradeoff price number connect airlin sourc deep web present challeng largescal integr optim access cost top dk queri web sourc unifi cost dbase approach studi problem support rank queri middlewar environ queri evalu multipl sourc studi web middlewar scenario queri web sourc motiv web travel agent scenario find restaur hotel exampl illustr real scenario benchmark queri experi access sourc capabl cost answer queri effici web middlewar coordin sourc sourc access incur network commun server comput paper aim optim access cost domin queri process relat dbm find top restaur chicago area highli rate close myaddr user rank queri sql syntax select order star addr myaddr stop queri queri answer middlewar access web sourc evalu predic intoscor aggreg score function return highest score restaur middlewar sourc queri answer figur show scenario evalu superpag capabl return score specif restaur random access return restaur descend order score sort access dinem similarli sort random access middlewar coordin access find top result access typic expens compar local comput vari cost character figur show averag access latenc includ network server time enabl ad dhoc rank data retriev data retriev find relev data larg databas problem user ami hous chicago search realtor constraint citi price rang bed bath return match hous ami realiz narrow queri extrem equal undesir hit manual oscil extrem eventu manag complet data retriev task relat databas offer support retriev task tradit boolean base queri model sql base hard criteria price user emploi soft criteria specif sens relev prefer unlik flat boolean result make holist schema match robust ensembl approach make holist schema match robust ensembl approach ranksql queri algebra optim relat top dk queri paper introduc ranksql system systemat principl framework support effici evalu rank top queri relat databas system rdbm extend relat algebra queri optim previous top queri process studi middlewar scenario rdbm piecem fashion focus specif oper sit core queri engin contrast aim support rank class databas construct kei insight rank relationship view logic properti data parallel membership properti relat data model membership essenti support rdbm support rank lack address fundament integr rank rdbm similar membership boolean filter support extend relat algebra propos rank relat model captur rank properti introduc extend oper support rank class construct enabl extend algebra present pipelin increment execut model rank queri plan express tradition base fundament rank principl optim top queri propos dimension enumer algorithm explor extend plan space enumer plan dual dimens rank membership propos sampl base method estim cardin rank awar oper cost plan experi show valid framework accuraci propos estim model ranksql queri algebra optim relat top dk queri paper introduc ranksql system systemat principl framework support effici evalu rank top queri relat databas system rdbm extend relat algebra queri optim previous top queri process studi middlewar scenario rdbm piecem fashion focus specif oper sit core queri engin contrast aim support rank class databas construct kei insight rank relationship view logic properti data parallel membership properti relat data model membership essenti support rdbm support rank lack address fundament integr rank rdbm similar membership boolean filter support extend relat algebra propos rank relat model captur rank properti introduc extend oper support rank class construct enabl extend algebra present pipelin increment execut model rank queri plan express tradition base fundament rank principl optim top queri propos dimension enumer algorithm explor extend plan space enumer plan dual dimens rank membership propos sampl base method estim cardin rank awar oper cost plan experi show valid framework accuraci propos estim model metaqueri queri structur web sourc dthe dfly recent wit rapid growth preval databas web recent survei april estim onlin databas deep web myriad onlin databas provid dynam queri base data access ranksql support rank queri relat databas manag system light dweight domain dbase form assist queri web databas fly queri rout find wai maze deepweb paper present sourc select system base attribut occurr framework rank select deep web sourc provid inform relev user requir huge number heterogen deep web data sourc end user sourc satisfi inform select rank sourc relev user requir challeng system find sourc user allow input imprecis initi queri kei insight observ semant relationship deep web sourc reveal queri interfac essenc occurr attribut base insight design occurr base attribut graph captur relev attribut rank sourc order relev user requir present iter algorithm realiz model preliminari evalu real world sourc demonstr effect approach optim multimod fusion multimedia data analysi organ structur web sourc queri schema cluster approach recent year web rapidli deepen preval databas onlin deep web sourc structur provid structur queri interfac result organ structur sourc domain hierarchi critic step integr heterogen web sourc observ structur web sourc queri schema attribut queri interfac discrimin repres sourc exploit sourc character view queri schema type categor data abstract problem sourc organ cluster categor data approach hypothes homogen sourc character hidden gener model schema find cluster govern statist distribut propos object function model differenti emploi principl hypothesi test maxim statist heterogen cluster evalu hundr real sourc schemabas cluster accur organ sourc object domain book movi cluster web queri schema model differenti function outperform exist likelihood entropi context linkag hierarch agglom cluster algorithm mine complex match web queri interfac cluster structur web sourc schema dbase web rapidli deepen preval databas onlin discov complex match web queri interfac correl mine approach enabl inform integr schema match critic step discov semant correspond attribut heterogen sourc complex match common complex search space exist techniqu focu simpl match tackl challeng paper take conceptu approach view schema match correl mine task match web queri interfac integr myriad databas internet deep web queri interfac gener form complex match attribut group correspond book domain observ cooccurr pattern queri interfac reveal complex semant relationship group attribut tend present queri interfac posit correl contrast synonym attribut neg correl rare occur insight enabl discov complex match correl mine approach develop dcm framework consist data prepar dual mine posit neg correl final match select unlik previou correl mine algorithm focu find strong posit correl algorithm care posit neg correl subtleti neg correl due special import schema match lead introduct correl measur measur distinct propos previou work evalu approach extens result show good accuraci discov complex match knock door deep web integr web queri interfac introduct recent wit rapid growth preval databas web recent survei decemb estim deep web sourc deep web myriad onlin databas provid dynam queri base data access queri interfac static url link door deep web essenti integr queri interfac integr deep web goal metaqueri project http metaqueri cs uiuc aim open deep web user build system user explor integr deep web sourc start focu integr deep web sourc domain book airfar import integr task typic scenario includ purchas book lowest price book sourc flight ticket trade price number connect airl understand web queri interfac deffort pars hidden syntax recent web rapidli deepen searchabl databas onlin data hidden queri form model integr web databas challeng understand queri interfac queri capabl sourc support automat extract interfac semant challeng queri form creat autonom approach build observ myriad sourc queri form reveal concert structur share common build block insight hypothes exist hidden syntax guid creation queri interfac albeit sourc hypothesi effect transform queri interfac visual languag prescrib grammar semant understand pars problem paradigm enabl principl solut declar repres common pattern deriv grammar systemat interpret queri form global pars mechan realiz paradigm address challeng hypothet syntax deriv secondari input heart form extractor develop grammar effort parser realiz pars mechan hypothet syntax experi show promis approach achiev accuraci extract queri condit random sourc knock door deep web integr web queri interfac introduct recent wit rapid growth preval databas web recent survei decemb estim deep web sourc deep web myriad onlin databas provid dynam queri base data access queri interfac static url link door deep web essenti integr queri interfac integr deep web goal metaqueri project http metaqueri cs uiuc aim open deep web user build system user explor integr deep web sourc start focu integr deep web sourc domain book airfar import integr task typic scenario includ purchas book lowest price book sourc flight ticket trade price number connect airl pebl web page classif neg exampl web page classif essenti techniqu web mine classifi web page interest class step mine web construct classifi interest class requir labori preprocess collect posit neg train exampl instanc order construct homepag classifi collect sampl homepag posit exampl sampl nonhomepag neg exampl collect neg train exampl requir arduou work caution avoid bia paper present framework call posit base learn pebl web page classif elimin manual collect neg train exampl preprocess pebl framework appli algorithm call map converg achiev high classif accuraci posit unlabel data high tradit svm posit neg data run stage map stage converg stage map stage algorithm weak classifi draw initi approxim strong neg data base initi approxim converg stage iter run intern classifi svm maxim margin progress improv approxim neg data class boundari eventu converg true boundari posit class featur space present algorithm support theoret experiment justif experi show set posit exampl algorithm outperform class svm accur tradit svm editori special issu web content mine phenomen growth web everincreas volum data inform publish numer web page research web mine aim develop techniqu effect extract mine knowledg mine semant larg scale integr web evid web rapidli deepen myriad searchabl databas onlin data hidden queri interfac larg scale integr deep web face challeng dynam ad hoc natur larg scale integr mandat dynam semant discoveri fly cope semant dynam discov sourc pre configur sourc specif knowledg tackl challeng initi work hing insight larg scale uniqu opportun observ desir semant connect surfac present characterist hidden regular sourc regular essenti leverag enabl semant discoveri report evid initi task integr deep web interfac extract schema match queri translat gener specif evid propos unifi insight mine semant larg scale integr exploit hidden regular holist sourc fulfil promis holist mine discuss challeng realiz dynam semant discoveri initi work relat effort wit unifi insight holist mine semant discoveri promis methodolog enabl larg scale integr structur databas web observ implic web rapidli deepen preval databas onlin potenti unlimit inform hidden queri interfac deep web searchabl databas import frontier data access paper survei unexplor frontier measur characterist pertin explor integr structur web sourc hand macro studi survei deep web larg april adopt random ip sampl approach million sampl larg deep web cover current directori servic hand micro studi survei sourc specif characterist sourc repres domain decemb hidden deep web sourc search engin cover data complex express queri form report observ publish result dataset research commun conclud implic necessarili subject shape research direct solut holist paradigm larg scale schema match schema match critic problem integr heterogen inform sourc tradition problem match multipl schema essenti reli find pairwis attribut correspond isol contrast propos match paradigm holist schema match match schema time find match handl set schema explor context inform reflect semant correspond attribut inform schema match pair realiz holist schema match develop altern approach global evalu local evalu global evalu exhaust assess model model express attribut match propos mg framework global evalu build hypothesi exist hidden schema model probabilist gener schema observ hand local evalu independ assess singl match increment construct model develop dcm framework local evalu build observ occurr pattern schema reveal complex relationship attribut appli approach match queri interfac deep web result show effect mg dcm approach demonstr promis holist schema match statist schema match web queri interfac schema match critic problem integr heterogen inform sourc tradition problem match multipl schema essenti reli find pairwis attribut correspond paper propos di erent approach motiv integr larg number data sourc internet deep web observ distinguish characterist er view schema match web scale ampl sourc provid structur inform domain book automobil sourc prolifer aggreg schema vocabulari converg small size motiv observ propos paradigm statist schema match unlik tradit approach pairwis attribut correspond holist approach match input schema find underli gener schema model propos gener statist framework mg hidden model discoveri consist hypothesi model gener select special gener framework develop algorithm mgssd target synonym discoveri canon problem schema match design discov model specif captur synonym attribut demonstr approach hundr real web sourc domain result show good accuraci heterogen learner web page classif classif interest class web page person homepag resum page interest problem typic machin learn algorithm problem requir class data train posit neg train exampl applic web page classif gather unbias sampl neg exampl appear difficult propos heterogen learn framework classifi web page elimin neg train data increas classif accuraci heterogen learner framework heterogen learner decis list linear separ complement elimin neg train data train phase increas accuraci test phase result show heterogen framework achiev high accuraci requir neg train data enhanc accuraci linear separ reduc error low margin data classifi accur requir human effort train pebl posit base learn web page classif svm web mine specif classifi web page user interest class step mine interest inform web construct classifi interest class requir labori pre process collect posit neg train exampl instanc order construct homepag classifi collect sampl homepag posit exampl sampl homepag neg exampl collect neg train exampl requir arduou work special caution avoid bias introduc paper posit base learn ing pebl framework web page classif elimin manual collect neg train exampl pre process present algorithm call map converg achiev classif accuraci posit unlabel data high tradit svm posit neg data experi show algorithm amount posit exampl tradit svm algorithm perform tradit svm minim probe support expens predic top dk queri paper address problem evalu rank top queri expens predic major dbmss support expens user defin predic boolean queri support rank queri import rank queri model user specif concept prefer relev similar call dynam user defin function middlewar system incorpor extern predic integr autonom sourc typic access object queri fuzzi join inher expens essenti user defin oper dynam associ multipl relat predic dynam defin extern access reli index mechan provid time sort output requir object probe evalu current standard sort merg framework rank queri effici handl predic complet probe object sort merg produc top answer minim expens probe develop formal principl probe determin probe absolut requir propos algorithm mpro implement principl provabl optim minim probe cost show mpro scale easili parallel experi real estat benchmark databas synthet dataset show mpro enabl signific probe reduct order magnitud faster standard scheme complet probe data mine web intellig data mine hold kei uncov databas research univers illinoi urbana dchampaign nbdl ci framework nsdl paper describ nbdl nation biologi digit librari project ci core integr system project nsf nsdl nation smete digit librari program approxim queri translat heterogen inform sourc paper present mechan approxim translat boolean queri constraint heterogen inform sourc achiev translat challeng sourc support constraint formul queri constraint precis translat instanc queri score perfectli translat rate site approxim grade unlik work gener framework adopt customiz close metric translat combin precis recal result show queri translat handl interdepend queri conjunct disjunct basi identifi essenti requir rule system user encod map atom semant unit algorithm translat complex queri rewrit term semant unit show approxim queri map account translat close paper present mechan approxim translat boolean queri constraint heterogen inform sourc achiev translat challeng sourc support constraint formul queri constraint precis translat instanc queri score gt perfectli translat rate gt site approxim grade unlik work gener framework adopt customiz close metric translat combin precis recal result show queri translat handl interdepend queri conjunct disjunct basi identifi essenti requir rule system user encod map atom semant unit algorithm translat complex queri rewrit term semant unit show practic assumpt algorithm gener approxim translat respect close metric choic present case studi show techniqu appli practic approxim queri translat heterogen inform sourc paper present mechan approxim translat boolean queri constraint heterogen inform sourc achiev translat challeng sourc support constraint formul queri constraint precis translat instanc queri score perfectli translat rate site approxim grade unlik work gener framework adopt customiz close metric translat combin precis recal result show queri translat handl interdepend queri conjunct disjunct basi identifi essenti requir rule system user encod map atom semant unit algorithm translat complex queri rewrit term semant unit show mind vocabulari queri map heterogen inform sourc paper present mechan translat constraint queri boolean express constraint heterogen inform sourc integr system difficult part wide rang constraint vocabulari formul queri describ algorithm appli user provid map rule translat queri constraint understood support context proper oper format show translat queri minim subsum origin translat queri compact unlik queri map work effect inter depend constraint handl constraint translat independ constraint fulli support framework explor relax semant rewrit closest support version sophist algorith powerbookmark system personaliz web inform organ user web overload inform medium pollut redund erron low qualiti inform www survei user conduct april user report find info problem user report organ collect inform problem empir studi user revisit pattern www page found individu page revisit studi result web user build organ larger collect bookmark futur refer maintain system architectur powerbookmark develop auspic larger webdb project address issu webdb hypermedia databas system built top nec percio object orient databas manag ment system unlik search engin fo predic rewrit translat boolean queri heterogen inform system refer field search predic specifi pattern match content field figur construct typic searchabl field ir system build index salton frake baeza yate faloutso direct search engin find document term word cat phrase joe doe index scheme field restrict queri gener wai index distribut object build stanford digit librari infobu extens constructor tool rapid describ extens constructor tool help inform expert librarian creat special queri synthes heterogen digit librari environ queri synthes graphic user interfac digit librari patron high level field multi sourc queri queri synthes interact queri translat attribut translat transform high level queri set sourc specif queri discuss constructor facilit discoveri attribut titl collat schema sourc select input widget synthes text box drop list widget support input control vocabulari design aspect describ prototyp constructor implement base stanford infobu metadata architectur keyword constructor tool queri synthes region schema queri gener queri transla conjunct constraint map data translat paper present mechan translat inform heterogen digit librari environ model inform set conjunct constraint satisfi real world object document metadata applic semant rule transform function constraint map understood support context machineri deal hierarch structur inform keyword constraint map data translat semant interoper introduct digit librari interact autonom system structur repres inform differ ensur semant interoper inform appropri map sourc context target context condit author john smith queri translat smith similarli compon date jan powerbookmark advanc web bookmark databas system inform share manag stanford infobu servic layer augment internet high dlevel inform manag protocol interoper digit librari worldwid evalu cost boolean queri map uniform queri languag make search heterogen inform sourc difficult approach user compos boolean queri rich front end languag user queri target sourc transform user queri subsum queri support sourc return extra document result process filter queri yield correct final result post filter approach involv signific cost document user retriev filter gener wai implement post filter batch post filter increment post filter paper evalu cost method search featur proxim oper experiment result show case increment post filter cost accept batch post filter cost extrem larg introduct emerg digi evalu cost boolean queri map uniform queri languag make search heterogen inform sourc difficult approach user compos boolean queri rich front end languag user queri target sourc transform user queri subsum queri support sourc return extra document result process filter queri yield correct final result post filter approach involv signific cost document user retriev filter gener wai implement post filter batch post filter increment post filter paper evalu cost method search featur proxim oper experiment result show case increment post filter cost accept batch post filter cost extrem larg introduct emerg digi metadata digit librari architectur design rational distribut heterogen proxi base digit librari autonom servic collect access indirectli proxi facilit metadata compat interoper digit librari design metadata architectur includ basic compon class attribut model proxi attribut model translat metadata facil search proxi metadata repositori attribut model proxi elev attribut set attribut defin class object relationship attribut captur attribut model translat map attribut attribut valu attribut model metadata facil search proxi provid structur descript collect search proxi provid access search capabl proxi final metadata repositori accumul select metadata local instanc start stanford propos internet meta dsearch experi paper stanford digit librari metadata architectur goal stanford digit librari project provid infrastructur afford interoper heterogen autonom digit librari servic servic includ search servic remot usabl inform process facil paper survei categor metadata requir divers set stanford digit librari servic built propos extens metadata architectur meet requir metadata architectur fit establish infrastructur promot interoper exist de facto metadata standard piec architectur implement construct architectur includ attribut model proxi attribut model translat servic metadata inform facil search servic local metadata repositori present discuss piec architectur show address motivati boolean queri map heterogen inform sourc search heterogen inform sourc difficult uniform queri languag approach user compos boolean queri rich front end languag user queri target sourc transform user queri subsum queri support sourc return extra document result process filter queri yield correct final result paper introduc architectur algorithm gener support subsum queri filter show gener subsum queri return minim number document discuss minim cost filter obtain implement prototyp version algorithm demonstr heterogen boolean system index term boolean queri queri translat inform retriev heterogen digit librari queri subsumpt filter introduct emerg digit libra 