SEASR Overview
-
Upload
loretta-auvil -
Category
Technology
-
view
904 -
download
0
description
Transcript of SEASR Overview
SEASR
National Center for Supercomputing Applications!University of Illinois at Urbana-Champaign
Loretta Auvil [email protected]
The SEASR project and its Meandre infrastructure!are sponsored by The Andrew W. Mellon Foundation
TheSEASRPicture
SEASROverview
SEASRwill:• helpscholarsaccessexis9nglargedatastoresmorereadily
• providescholarswithenhanceddatasynthesisandqueryanalysis– fromfocuseddataretrievalanddataintegra9on
– tointelligenthuman‐computerinterac9onsforknowledgeaccess
– toseman9cdataenrichment
– toen9tyandrela9onshipdiscovery– toknowledgediscoveryandhypothesisgenera9on
• empowercollabora9onamongscholarsbyenhancingandinnova9ngvirtualresearchenvironments
AQuickLookatSEASR
• Addresses:– Challengesoftransforminginforma9onintoknowledge
– Construc9ngthesoGwarebridgestomovefromtheunstructuredandsemi‐structureddataworldtothestructureddataworld.
• Aims:– Makedigitalcollec9onsmoreuseful– Provideaccesstorelevantanaly9csandvisualiza9ons
– EnableeasymashabilityviaSOA
SEASR:Reach+Relevance+Reuse+Repeatability
SEASRemphasizesflexibility,scalability,modularity,providescommunityhubandaccesstoheterogeneousdataandcomputa9onalsystems– Seman9cdrivenenvironmentforSOAinteroperability– Encouragessharingandpar9cipa9onforbuildingcommuni9es– Modularconstruc9onallowsflowstobemodifiedandconfiguredto
encouragereusabilitywithinandacrossdomains– Enablesamashupandintegra9onoftools– Data‐intensiveflowscanbeexecutedonasimpledesktoporalarge
cluster(s)withoutmodifica9on– Computa9oncanbecreatedfordistributedexecu9ononserverswhere
thecontentlives– Useraccessibilitytocontroltrustandcompliancewithrequiredcopyright
licenseofcontent– ReliesonstandardizedResourceDescrip9onFramework(RDF)todefine
componentsandflow
SEASRTextAnaly9csGoalsAddresstheScholarlytextanaly9csneedsby:
• EfficientlymanagingdistributedLiteraryandHistoricaltextualassets• Structuringextractedinforma9ontofacilitateknowledgediscovery• Extractinforma9onfromtextatalevelofseman9c/func9onal
abstrac9onthatissufficientlyrichtosupportques9on‐answering• Devisearepresenta9onfortheextractedinforma9onthatcanbe
efficientlyreasonedovertorecoverdataintheques9on‐answerprocess
• Devisealgorithmsforques9onansweringandinference• DevelopUIforeffec9vevisualknowledgediscoverywithseparate
querylogicfromapplica9onlogic• Leveragingexis9ngapproachesanddevisealgorithmsforclustering,
inference,andQ&A• DevelopinganInterac9onUIforeffec9vevisualdataexplora9on• Enablethetextanaly9csthroughSEASRcomponents
Workbench
• Web‐basedUI
• Componentsandflowsareretrievedfromserver
• Addi9onalloca9onsofcomponentsandflowscanbeaddedtoserver
• Createflowusingagraphicaldraganddropinterface
• Changepropertyvalues• Executetheflow
CommunityHub
SEASR@Work–Zotero
• PlugintoFirefox• Zoteromanagesthe
collec9on
• LaunchSEASRAnaly9cs– Cita9onAnalysisusestheJUNG
networkimportancealgorithmstoranktheauthorsinthecita9onnetworkthatisexportedasRDFdatafromZoterotoSEASR
– ZoteroExporttoFedorathroughSEASR
– SavesresultsfromSEASRAnaly9cstoaCollec9on
• LaunchMONKProcessing– MONKDBInges9onWorkflow
WebService
Interac9veWebApplica9on
SEASR@Work–Fedora
SEASR@Work–En9tyMash‐up
• En9tyExtrac9onwithOpenNLP
• Loca9onsviewedonGoogleMap
• DatesviewedonSimileTimeline
SEASR@Work–AudioAnalysis• NEMA:ExecutesaSEASR
flowforeachrun
– Loadsaudiodata– Extractsfeaturesforevery
10secmovingwindowofaudio
– Loadsandappliesthemodels
– SendsresultsbacktotheWebUI
• NESTER:Annota9onofAudioviaSpectralAnalysis
SEASR@Work–MONK
Executesflowsforeachanalysisrequested– Predic9vemodelingusingNaïveBayes
– Predic9vemodelingusingSupportVectorMachines(SVM)
SEASR@Work–DISCUS• On‐demandusageof
analy9cswhilesurfing– Whilenaviga9ng
requestanaly9cstobeperformedonpage
– Textextrac9onandcleaning
• Summariza9onandkeyworkextrac9on– Listtheimportant
termsonthepagebeinganalyzed
– Providerelevantshortsummaries
• Visualmaps– Provideavisual
representa9onofthekeyconcepts
– Showthegraphofrela9onsbetweenconcepts
SEASRandUIMA:Emo9onTrackingGoalistohavethistypeofVisualiza9ontotrackemo9onsacrossatextdocument(Leveragingflare.prefuse.org)