Linked Data: Now What?
         Aldo Gangemi
    STLab, ISTC-CNR, Rome
      aldo.gangemi@cnr.it



                      ...
What was right
• “Just do it” bootstrapping for the Semantic Web
• Few simple recipes for reengineering, publishing,
  ali...
Issues
1. Sparseness of data (unless controlled by a schema)
2. Problems of evolution and versioning (both on schema and d...
What should(n’t) we do
• Many problems (e.g. 1 through 4) will be solved
  as a natural evolution of the technology
• Some...
Some design directions
• What domain semantics is piped into reengineered linked datasets?
   – are bridging approaches (e...
Prochain SlideShare
Chargement dans…5
×

Aldo Gangemi's view on the future of Linked Data

765 vues

Publié le

Aldo Gangemi's presentation for the ESWC 2010 Panel entitled "Linked Data: Now what?"

Publié dans : Technologie, Formation
0 commentaire
0 j’aime
Statistiques
Remarques
  • Soyez le premier à commenter

  • Soyez le premier à aimer ceci

Aucun téléchargement
Vues
Nombre de vues
765
Sur SlideShare
0
Issues des intégrations
0
Intégrations
7
Actions
Partages
0
Téléchargements
0
Commentaires
0
J’aime
0
Intégrations 0
Aucune incorporation

Aucune remarque pour cette diapositive





  • Aldo Gangemi's view on the future of Linked Data

    1. 1. Linked Data: Now What? Aldo Gangemi STLab, ISTC-CNR, Rome aldo.gangemi@cnr.it 1
    2. 2. What was right • “Just do it” bootstrapping for the Semantic Web • Few simple recipes for reengineering, publishing, aligning, and consuming data • Refreshing attention to practical problems, technologies that scale, and to the webby side of the Semantic Web • URI-based data integration proven feasible • Concrete platform and use case for a viral effect in opening data (cf. biology, gov data) 2
    3. 3. Issues 1. Sparseness of data (unless controlled by a schema) 2. Problems of evolution and versioning (both on schema and data sides) 3. Licensing and policies partly unclear 4. Current topic coverage is scattered in extension and depth 5. Difficulty of exploring data just to know what they are about (exceptions, e.g. RelFinder) 6. Lack of good interaction with linked data 7. What recipes for data created from inference, enrichment, lenses, customized consumption? 8. Semantics of many datasets is suspicious 9. Much knowledge is in literals rather than entities 3
    4. 4. What should(n’t) we do • Many problems (e.g. 1 through 4) will be solved as a natural evolution of the technology • Some problems (e.g. 5 through 7) are common to all semantic technologies: are semantic data special from an interaction viewpoint? • Some problems (e.g. 8/9) depend on limited attention to design aspects • Let’s stand up to the SW vision and to the interdisciplinarity of Web Science • Please don’t reinvent the wheel 4
    5. 5. Some design directions • What domain semantics is piped into reengineered linked datasets? – are bridging approaches (e.g. D2R, Virtuoso Sponger cartridges) sustainable in presence of legacy data or specific requirements? – e.g. Freebase Gridworks allows some customization when reengineering DBs – e.g. OPPL tool from University of Manchester allows pattern-based refactoring – e.g. Semion tool from STLab keeps track of, and enables custom semantic transformations when reengineering • At the carrefour between informal and formal semantics – extensive usage of metamodels: SKOS, Lexical, DB, etc. – possibilities from OWL2 punning mechanism – heterogeneous ontology matching techniques can help with semantic conflicts and with the literal vs. entity issue – check next workshops: KIELD@EKAW2010 and WOP@ISWC2010 – check Ontology Alignment Evaluation Initiative on automatic data interlinking • LOD meets Ontology Design – good practices and design patterns (just started, but more communication is needed) – OWL LOD datasets with task-oriented ontologies – e.g. Semantic Scout application (paper@EKAW2010) 5

    ×