Diese Präsentation wurde erfolgreich gemeldet.
Wir verwenden Ihre LinkedIn Profilangaben und Informationen zu Ihren Aktivitäten, um Anzeigen zu personalisieren und Ihnen relevantere Inhalte anzuzeigen. Sie können Ihre Anzeigeneinstellungen jederzeit ändern.

2014 06-04-presentation-mdn-2014

243 Aufrufe

Veröffentlicht am

Presentation to the Metadata Developer Network Workshop 2014 (MDN Workshop 2014), 4th of June 2014, Geneva, Switzerland.

Also available on http://www.slideshare.net/dri_ireland

Abstract: In this presentation, we report on our experience using the EBU Core OWL ontology for annotating audiovisual archival content stored in an EBU Core XML Schema compliant tool used by RTÉ, the national public service broadcaster of the Republic of Ireland. We first describe the goal of the project and elaborate on the role of Semantic Web ontologies and technologies. We continue with a report on some of the challenges while using the EBU Core OWL ontology. We finally formulate several recommendations on the conceptual model and the ontology development method.

Veröffentlicht in: Technologie, Bildung
  • Als Erste(r) kommentieren

  • Gehören Sie zu den Ersten, denen das gefällt!

2014 06-04-presentation-mdn-2014

  1. 1. RTÉ Content Discovery Project Christophe Debruyne c.debruyne@ria.ie christophe.debruyne@insight-centre.orgchristophe.debruyne@insight-centre.org MDN Workshop -- 4th of June 2014
  2. 2. Outline • Context • Goal and Challenges of the RTÉ Content Discovery Project • Tasks and Data Annotation• Tasks and Data Annotation • EBU Core – Identification of problems • Addressing the issues • Using the ontology • Conclusions and Recommendations
  3. 3. Context RTÉ, Ireland's National Television and Radio Broadcaster National trusted digital repository for Ireland's social and cultural data. Centre for Data Analytics Documents Television Radio Stills Linking and preserving data held by Irish Institutions with central internet access point. • Standards • Cataloguing • Archiving • Preservation • Insight @ NUIG = DERI • Semantic Technologies • Linked Data • Data Analytics Platform
  4. 4. Goal of the RTÉ Content Discovery Project • Discover implicit knowledge • across the different archives • and the Web of Data • To facilitate internal workflows (e.g., search) • For wider reuse and repackaging RTÉ’s Documents Television Radio Stills • For wider reuse and repackaging RTÉ’s information • Challenges • Heterogeneous databases • Different guidelines and practices • Legacy data (from previous systems) • … “Linking Open Data cloud diagram, by R. Cyganiak and A. Jentzsch. http://lod-cloud.net/”
  5. 5. Part of a wider ambition …
  7. 7. RTÉ Content Discovery In this presentation we focus on Television and Radio archives Documents Television Radio Stills • In this presentation we focus on Television and Radio archives • The Television and Radio archives • Are maintained on two different instances of the same system • A system that is EBU Core “compatible” • Different content, different guidelines, …
  8. 8. Three main tasks • Annotate the data. • Using relevant standards, ontologies and vocabularies. • Resource Description Framework (RDF). • Obtain an integrated view of the different archives by creating links between the RDF representations of RTÉ’s archival assets across the different archives. • Apply advanced methods for discovering related data for a given subject in external sources such as the Linked Data Cloud.
  9. 9. Data annotation Relational Database D2RQ RDF Dump Triplestore Television Radio Map symbols of database to predicates (relations and concepts) in chosen ontologies / vocabularies Use D2RQ to generate RDF dump Store RDF dump in adequate triple store (Jena TDB) Which ontologies? • Dublin Core, DC Terms • Foaf • EBU Core OWL • …
  10. 10. EBU Core OWL • The RTÉ Content Discovery platform will rely on Semantic Web technologies to reason. Ontologies will therefore need to be correct. • But … while adopting the EBU Core OWL ontology, several problems where identified. • We contacted EBU to resolve these issues. • We provide an overview of some of these problems.
  11. 11. Problems • (1) Forgotten concept unions • The property ebucore:description has multiple domain axioms. <rdfs:domain rdf:resource="&ebu;BusinessObject"/> <rdfs:domain rdf:resource="&ebu;MediaResource"/> • Unintentionally the wrong implicit information can be inferred.• Unintentionally the wrong implicit information can be inferred. • (2a) Property unsatisfiability – via class axioms <owl:Class rdf:about="&ebu;BusinessObject"> … <owl:disjointWith rdf:resource="&ebu;Resource"/> … </owl:Class> • Because of (1) and (2), the property description could not be used
  12. 12. Problems • (2b) Property unsatisfiability – role hierarchies and datatypes • Duration has the range xsd:string • The subproperties of duration have other ranges (e.g., double in the case of duration in edit units) • Because each subproperty also inherits the range of the• Because each subproperty also inherits the range of the superproperty, all instances in the object of that property must be at the same time a string, and a double. This type conflict results in a contradiction. • With (2a) and (2b) we identified 40 properties that lead to problems.
  13. 13. Problems • (3) Inconsistencies between formal and informal definitions • BusinessObject is defined as: "An image, a document, an annotation […], a tag […], or an audiovisual media resource […]. Other types of BusinessObjects may be defined as subclasses.“BusinessObjects may be defined as subclasses.“ • Resource is defined as: "A manifestation of a BusinessObject." and disjoint with BusinessObjects. Meaning no individual can be an element of BusinessObjects and Resources at the same time. • The domain of a title is BusinessObject, yet, it’s definition is: "Specifies the title or name given to the resource. […]"
  14. 14. Problems • (4) User readable labels • Many different properties have the same human readable label, which could confuse the end user – e.g., when generating an Interface. • E.g., there were 11 properties with the label “Name”• E.g., there were 11 properties with the label “Name” • Some properties had empty labels • (5) Roles – Loss of context • Agents were related to Business Objects (BO) • Agents were related to a Role • But … a role did not relate to agents in relationship with a BO • This lead to a loss of context.
  15. 15. Addressing the issues • Problems were addressed over email. • The discussions are “lost”, traces are only known to us … • The ontology-engineering activities of EBU Core should adopt appropriate methods and tools for collaboration. • Participation of others • Traceability (!) • The ontology is still being developed as we go along, and we have been able to make (parts of it) work…
  16. 16. Using the ontology
  17. 17. Conclusions and Recommendations • RTÉ Archives aims at a wider reuse and repackaging of their archival content on digital platforms through the innovative use of Semantic and Linked Data technologies. • We adopted the EBU Core OWL ontology for annotating the television and radio archives, yet identified some issues. • We adopted the EBU Core OWL ontology for annotating the television and radio archives, yet identified some issues. • We collaborated on resolving those issues together with EBU • However, we feel that appropriate collaborative methods and tools should be adopted to facilitate the ontology- engineering process and – more importantly – enable other to participate AND have visible traceability of the decisions.
  18. 18. References • D2RQ, http://d2rq.org/ • Digital Repository of Ireland, http://www.dri.ie/ • Insight, http://www.insight-centre.org/• Insight, http://www.insight-centre.org/ • Jena TDB, http://jena.apache.org/documentation/tdb/ • RTÉ Archives, http://www.rte.ie/archives