answerart contextu question answer real dtime new recommend system social network analysi ontolog edit log visual web page content semant technolog social network analysi ontolog edit log machin learn knowledg discoveri databas knowledg discoveri databas nontrivi extract implicit previous unknown potenti inform databas argu contrast machin learn knowledg discoveri databas appli real world databas real world databas larg rais problem access real world databas access databas manag system number access reduc minimum properti forc standard set orient interfac relat databas manag system order appli method knowledg discoveri databas present system discov data depend build set orient interfac point main effort put discoveri restrict unari inclus function depend relat databas system embodi infer rel model common real dword relat tripl extract dgram entiti resolut text statist learn ontolog enhanc web page content visual firefox visual ontobridg semi dautomat semant annot softwar semi dautomat categor video videolectur net demo historyviz visual event relat extract wikipedia semant web wiki social network analysi page edit document visual base semant graph visual analysi document semant graph construct event templat written new semant technolog captur commun insid organ semant graph deriv triplet applic document summar guest editor introduct special issu ecml pkdd predict categori addit topic hierarchi monitor network evolut mdl abstract public titl author evolut scientif topic commun time commun shrunk emerg split time time turn point propos timefal automat answer question social network graph evolv time main novelti propos approach user defin paramet reli principl minimum descript length mdl extract commun find good cut point time commun chang abruptli cut point good lead shorter data descript illustr algorithm synthet larg real dataset show result timefal agre human intuit cross dlingual search european languag rippl rule learn autom word lemmatis featur select classif larg document collect machin learn ecml machin learn ecml ontogen semi dautomat ontolog editor abstract paper present semi automat ontolog editor implement version ontogen system system integr machin learn text mine algorithm effici user interfac lower entri barrier user profession ontolog engin main featur system includ unsupervis supervis method concept suggest concept name ontolog concept visual system test extens user trail realworld scenario posit result knowledg discoveri databas pkdd web social web discov deploi user content profil ontogen semi dautomat ontolog editor abstract paper present semi automat ontolog editor implement version ontogen system system integr machin learn text mine algorithm effici user interfac lower entri barrier user profession ontolog engin main featur system includ unsupervis supervis method concept suggest concept name ontolog concept visual system test extens user trail realworld scenario posit result text mine link analysi softwar mine text mine link analysi softwar mine extract name entiti relat time base wikipedia paper present approach mine inform relat peopl place organ event extract wikipedia link time scale approach consist phase identifi relev page categor articl peopl place organ gener timelin link name entiti extract event time frame illustr propos approach million wikipedia articl povzetek predstavljen metod rudarjenja informacij iz wikipedi urejanj asovno zgradbo semant address issu semant agent commun languag compar combin approach autom subject classif text abstract machin learn string match approach autom subject classif text compar perform advantag downsid approach base svm algorithm compris string match control vocabulari word text classifi data collect consist subset compendex classifi class shown svm averag outperform string match approach hypothesi svm yield recal string match precis confirm class approach complementari investig combin base combin vocabulari result shown origin approach machin learn approach background knowledg control vocabulari string match approach base control vocabulari outperform approach combin automat manual obtain term reason result investig includ larger data collect combin predict background knowledg ontolog construct knowledg ontolog construct user profil web abstract brows web provid profil inform social network servic tag pictur user leav plethora trace paper analyz natur trace investig user data distribut web system examin wai aggreg user profil inform analys focu explicitli provid profil inform homepag activ data tag assign bookmark imag experi reveal signific benefit interweav profil inform complet profil advanc foaf vcard profil gener disclosur facet user higher level inform induc profil higher precis predict tag base profil solv cold start problem semi dautomat construct topic ontolog knowledg ontolog construct string kernel classif slovenian web document appli collabor filter real dlife corpor data text classif activ learn paper show text classifi label train data reduc combin activ learn expect maxim em pool unlabel data queri committe activ select document label em naiv bay model improv classif accuraci concurr estim probabilist label remain unlabel document improv model present metric measur disagr committe member account strength disagr distribut document experiment result show method combin em activ learn requir half label train exampl achiev accuraci em activ learn keyword text classif activ learn unsupervis learn inform retriev introduct set learn text classifi obtain lab text mine action featur select dimension reduct find reliabl meaning pattern data high number attribut extrem difficult featur select help decid attribut combin attribut import find pattern chapter studi featur select method build classif model high throughput genom microarrai proteom mass spectrometri data set thousand featur candid analyz compar combin data set describ basic approach featur select illustr effect ms cancer proteom data set close discuss assist perform analysi high dimension genom proteom data data sparsiti issu collabor filter framework abstract amount inform web grow rapidli dai automat filter inform order ensur greater user effici emerg field user profil web person popular content filter techniqu develop chapter present techniqu collabor filter give overview collabor filter approach present experiment result confront nearest neighbor knn algorithm support vector machin svm collabor filter framework dataset properti nearest neighbor algorithm collabor filter task support vector machin consid state art classif algorithm collabor filter interpret classif regress task virtual supervis learn algorithm svm appli experi perform standard publicli dataset hand real life corpor dataset fit profil ideal data collabor filter conclud qualiti collabor filter recommend highli depend sparsiti data show knn domin dataset low sparsiti svmbase approach perform highli spars data workshop twentieth nation confer artifici intellig simpl classif larg topic ontolog web document abstract paper conceiv summari previou work explor altern learn model classif document structur output provid discuss strong weak point method visual text document corpu visual commonli data analysi user initi idea raw data visual represent regular obtain analysi similar talk autom text process data consist text document visual text document corpu autom text process point view natur languag redund sens word share common similar mean comput hard understand background knowledg describ approach visual text document collect base method linear algebra appli latent semant index lsi techniqu help extract background knowledg corpu text document view extract hidden semant concept text document visual help data analysi instanc find main topic larger set document extract main concept document techniqu lsi make result visual set descript european research project fp find main area project cover includ semant web learn secur paper describ method visual document corpu base lsi system implement give result system dataset povzetek predstavljena je vizualizacija korpusa besedil build minor languag corpora learn gener web search queri web obviou sourc valuabl inform process collect organ util resourc difficult describ corpusbuild approach automat gener web search queri collect document match minor concept concept text document belong minor natur languag web individu document automat label relev relev languag filter feedback learn queri length inclus exclus term select method help find previous unseen document target languag system learn select good queri term varieti term score method find odd ratio score calcul document acquir consist accur queri gener method parameter queri length gamma distribut present empir result learn method vari time horizon learn result past queri find system perform initi document hand word elicit user experi appli approach multipl languag present show approach gener languag initi condit kdd workshop report link discoveri issu web mine web semant web visual larg graph cluster neighborhood abstract paper present method visual larg graph dimension space collect web page main contribut represent chang enabl handl data idea method consist major step transform graph spars matrix vertex graph spars vector matrix spars vector compon vertic close vertex repres vector perform hierarch cluster hierarch mean set spars vector result hierarchi cluster step map hierarchi cluster dimension space similar clu ster close pictur effect procedur assign uniqu coordin vertex vertic group vert ic level hierarchi stronger connect graph place closer pictur method power distribut graph show applic method real world exampl visual institut collabor graph cro ss sell recomm ndation graph featur select linear classifi weight interact classif model paper explor featur score select base weight linear classif model investig method combin learn model compar analysi includ learn algorithm na ve bay perceptron support vector machin svm combin featur weight method odd ratio inform gain weight linear model linear svm perceptron experi show featur select weight linear svm yield classif perform featur weight method combin explor learn algorithm result support conjectur sophist featur weight method appar compat learn algorithm improv classif perform kdd workshop report link analysi group detect linkkdd roadmap web mine web semant web map document web page ontolog featur select hierarchi web document learn word normal word suffix context unlabel data model inform textual data combin label unlabel data mine web creat minor languag corpora web valuabl sourc languag specif resourc process collect organ util resourc difficult describ corpusbuild approach automat gener web search queri collect document minor languag differ pseudo relev feedback retriev document label automat languag classifi relev irrelev feedback gener queri experi queri gener method queri length find inclus exclus term help retriev document target languag find odd ratio score calcul document acquir consist accur queri gener method describ experi hand word elicit user initi document show method perform similarli experi appli approach multipl languag present show approach gener varieti languag automat web search queri gener creat minor languag corpora paper describ techniqu requir user give hand keyword document automat collect languag specif resourc web present system automat gener web search queri construct corpora minor languag propos approach requir human intervent system provid freeli languag filter initi document target languag order magnitud effici fit crawler languag filter onlin learn web queri gener find document match minor concept web paper describ approach learn gener web search queri collect document match minor concept case studi concept text document belong slovenian minor natur languag web individu document automat label relev relev languag filter feedback learn queri length inclus exclus term select method help find previous unseen document target languag system corpusbuild learn select good queri term varieti term score method present empir result learn method vari time horizon learn result past queri approach gener languag initi condit data mine basket collect locat year text mine workshop session titl text mine integr relat research area report kdd workshop text mine paper give overview kdd workshop text mine held boston ma august report detail research issu cover paper present workshop group discuss held final session workshop keyword text mine inform extract inform retriev natur languag process kdd workshop report introduct grow import electron media store dissemin text document creat burn tool techniqu assist user find extract relev inform larg data repositori inform manag organ maintain structur databas focu data mine research emerg world wide web extend focu mine inform unstructur semi structur inform sourc line featur select unbalanc class distribut naiv bay paper describ approach featur subset select take account problem specif learn algorithm characterist develop naiv bayesian classifi appli text data combin address learn problem focu domain featur highli unbalanc class distribut asymmetr misclassif cost implicitli problem asymmetr misclassif cost class valu target class predict prefer fals posit fals neg problem automat document categor machin learn identifi document relev select categori exampl belong select categori experiment comparison eleven featur score measur show domain algorithm characterist significantli impro turn yahoo automat web dpage classifi paper describ approach automat web page classif base yahoo hierarchi machin learn techniqu develop learn text data hierarch classif structur high number featur reduc take account hierarch structur featur subset select base method inform retriev document repres featur vector includ gram includ singl word unigram commonli learn text data base hierarch structur problem divid subproblem repres categori includ yahoo hierarchi result learn set independ classifi predict probabl member categori experiment evalu real world data show propos approach good result half test featur subset select text dlearn paper describ method featur subset select larg text data experiment comparison real world data collect web user show characterist problem domain machin learn algorithm consid featur score measur select problem domain consist hyperlink form small document repres word vector learn experi naiv bayesian classifi text data perform achiev featur select method base featur score measur call odd ratio inform retriev brows dassistenten tour guid und adapt www dserver machin learn outcom predict patient sever head injuri machin learn techniqu interpret result discret event simul combinatori optim induct concept learn visual domain concept descript stochast search induct logic program abstract stochast search space candid claus appeal scale ilp larg dataset address approach bayesian network model adapt guid search space examin guid search area previous perform area ilp explor show improv area curv recal precis curv modif 