publishing and interacting with linked data

30
Publishing and Interacting with Linked Data Roberto Garcia, Josep Maria Brunetti, Antonio López-Muzás, Juan Manuel Gimeno, Rosa Gil WIMS’11 Conference, May 25 th 2011, Sogndal, Norway HCI & Data Integration Research Group Universitat de Lleida, Spain

Upload: kuniko

Post on 11-Jan-2016

28 views

Category:

Documents


1 download

DESCRIPTION

Publishing and Interacting with Linked Data. Roberto Garcia, Josep Maria Brunetti, Antonio López-Muzás, Juan Manuel Gimeno, Rosa Gil WIMS ’ 11 Conference, May 25 th 2011, Sogndal, Norway. HCI & Data Integration Research Group Universitat de Lleida, Spain. Overview. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Publishing and Interacting with  Linked Data

Publishing and Interacting with Linked Data

Roberto Garcia, Josep Maria Brunetti, Antonio López-Muzás, Juan Manuel Gimeno, Rosa Gil

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

HCI & Data Integration Research GroupUniversitat de Lleida, Spain

Page 2: Publishing and Interacting with  Linked Data

Overview

• RhizomerPlatform for Linked Data publishing (also Semantic Data)–Data access

• Read (LinkedData / SPARQL) • Write (Semantic FORMS + HTTP PUT/POST/DEL)

–Data sources• Jena (File, MySQL, Postgres), Virtuoso, OWLIM,…

–Interaction• Overview: Navigation bars• Filter: Facets• Details: HTML+RDFa, Interaction Web Services

(map, timeline,…)

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 3: Publishing and Interacting with  Linked Data

Architecture

• Server part:– REST interface over repository

• GET: URI (LinkedData) or SPARQL Query• PUT/POST/DELETE: data management

MetadataStore

Rhizomer

GE

T

PU

T

PO

ST

DE

L

search new edit delete

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 4: Publishing and Interacting with  Linked Data

Architecture

• Scope of GET operations:– Concise Bounded Description (CBD) +

rdfs:labels for all selected resources

• Scope of POST/PUT/DEL operations:– Concise Bounded Description

• DESCRIBE queries by default– Facilitate browsing heterogeneous data– Without a priori knowledge

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 5: Publishing and Interacting with  Linked Data

http://rhizomik.net/~rosa

http://...vcard#ADR

http://...vcard#StreetJaume II, 69

http://...vcard#Locality

Lleida

http://...onto#affiliation

http://www.udl.cat

http://...rdf#type

http://...rdfs#label

Rosa Gil

http://...rdf#type

http://...onto#Professor

http://...rdf#type

http://...vcard#work

http://...onto#University

Architecture

• DESCRIBE http://rhizomik.net/~rosa: CBD(…~rosa) = Fragment A U rdfs:labels

Fragment A

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 6: Publishing and Interacting with  Linked Data

Architecture

RDF2HTML

GET RDF2HTML

<s5t:Audio rdf:about="http://www.segre.com/audio/20070323-bm.mp3">      <dc:title>Butlletí Migdia</dc:title>      <dc:date>2007-03-23</dc:date>      <s5t:genre rdf:resource="&srs;11000000"/>      <s5t:transcriptResource>              http://www.segre.com/audio/20070323-bm.xml       </s5t:transcriptResource></srt:Audio> http://www.segre.com/ audio/ 20070323-bm.mp3 a Audio

title Butlletí Migdia

date 2007-03-23

genre politics

transcript http://www.segre.com/audio/20070323-bm.xml

Referrers

http://rhizomik.net/.../NewsCodes-SubjectsOnto#11000000 a Subject

label politics

descriptionLocal, regional, national and international exercise of power, or struggle for power, and the relationships between governing bodies and states.

Referrers

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 7: Publishing and Interacting with  Linked Data

Interacting

• Until now, Rhizomer like other Linked Data publishing tools…– HTML+RDFa and content negotiation– Useful for computers, but also for human users?

• User tests:– Typical questions:

• Where do I start? • Where do I go now?• What is this data about?

– What do we offer? • Text search, type URI, SPARQL query,…

…but they do not answer end-users needs

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 8: Publishing and Interacting with  Linked Data

Interacting

• Example: What to do with DBPedia? – 3.5 million things described

• Ontology: 257 classes y 1276 properties

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 9: Publishing and Interacting with  Linked Data

Interacting

– DBPedia main page does not provide too many clues...

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 10: Publishing and Interacting with  Linked Data

Interacting

• SPARQL Endpoint:– select distinct(?c) (count(?i) as ?n)

where {?i a ?c} order by desc(?n)

c n

http://www.w3.org/2002/07/owl#Thing 1668503

http://www.w3.org/2004/02/skos/core#Concept 632607

http://www.opengis.net/gml/_Feature 571764

http://dbpedia.org/ontology/Place 462349

http://dbpedia.org/ontology/Person 363751

http://dbpedia.org/ontology/Work 355100

http://dbpedia.org/ontology/PopulatedPlace 340443

http://dbpedia.org/ontology/Settlement 304370

http://xmlns.com/foaf/0.1/Person 296595

http://dbpedia.org/ontology/Species 168575WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 11: Publishing and Interacting with  Linked Data

Interacting

• Simple search– What to type?

– A URI? URI label?

– …

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 12: Publishing and Interacting with  Linked Data

Interacting

• Advanced search:– iSparql

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 13: Publishing and Interacting with  Linked Data

Information Architecture

• Proposal: adapt data interaction patterns [Shneiderman] and Information Architecture components [Morville]– Data structure: automate IA components

– Scalable, dataset independent

• Interaction Patterns / IA Components– Overview --> navigation bars, site map

• Main classes (topics) and their organisation

– Zoom & Filter--> facets, breadcrums• Main properties and values, filter

– Details --> HTML view, Object-Action Paradigm

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 14: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

– Get hierarchical structure for dataset ontologies

– For each class• URI

• Label

• # instances

• Subclasses list

– Flatten to desired # levels and elements / level• Classes and namespaces whitelist and blacklist

– Algorithm:• When there is room, divide class with most instances

• When too many options, group classes with less instances

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 15: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

7 elements in “Species” submenu:

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 16: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 17: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 18: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 19: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 20: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 21: Publishing and Interacting with  Linked Data

AI Components Navigation Bars

Navigation bar provides overview for DBPedia… …but what to do with 12.334 birds now?

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 22: Publishing and Interacting with  Linked Data

AI Components Facets

• Pre-compute list of facets for each class– Properties used in dataset instances descriptions– Facet metrics: frequency, #different values, more

common value cardinality, entropy,…

• DBPedia Birds case:–  http://dbpedia.org/ontology/Bird 12334 instances– 226 different properties

• http://dbpedia.org/ontology/kingdom, 100%, 3 values, 6846 (Animalia),…

• User unfolds facet, show 5 +common values– Filter, get more…

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 23: Publishing and Interacting with  Linked Data

AI Components Facets

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 24: Publishing and Interacting with  Linked Data

AI Components Facets

• Order facets by "utility” (work in progress):– Prefer common properties, high frequency

– For values cardinality, discard extreme cases: • Just 1 value for the property

• All different values, i.e. max. value cardinality = 1

– Prefer properties that evenly divide results, higher entropy

Property Entropy Value InstanceshasFlavor 0.918434 Moderate 28

Strong 15Delicate 10

hasSugar 0.556416 Dry 43Sweet 6OffDry 4

Wine (Wine Ontology)

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 25: Publishing and Interacting with  Linked Data

InteractionObject-Action

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 26: Publishing and Interacting with  Linked Data

InteractionObject-Action

• Given a resource (object)…determine available services (actions)

• Services descriptions:– endPoint: service location– hasInput: ASK SPARQL Query Template

• Service applicable to any resource ASK is true

– hasOutput: output content type

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 27: Publishing and Interacting with  Linked Data

InteractionObject-Action

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 28: Publishing and Interacting with  Linked Data

Conclusions

• Navigation menus – dataset classes (topics) overview

• Facets – per class properties and values overview, filter

• Details – Generic view (HTML) + specific (map, timeline…)

• Users build queries without SPARQL or dataset structure knowledge

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 29: Publishing and Interacting with  Linked Data

Future Work

• Facets: – type dependent facets (number ranges, dates,…),

inverse properties, facets for facet values,…

• Other IA components: sitemap, breadcrumbs,…

• Assisted metadata edition– Recommend properties and values based on

available ontologies and values

• More interaction services– Calendar– Plot…

WIMS’11 Conference, May 25th 2011, Sogndal, Norway

Page 30: Publishing and Interacting with  Linked Data

Thank you for your attention

More details: http://rhizomik.net/rhizomer

Contact: Roberto Garcia ([email protected])

HCI & Data Integration Research GroupUniversitat de Lleida, Spain