WWW 2013 - Linked Data at Globo.com

Transcript

1. Linked Data at globo.com Semantic Team semantica@corp.globo.com Ícaro Medeiros globo.com Web of Linked Entities 2013 WWW 2013
2. Who we are
3. BROADCAST MOVIES PAY TV INTERNET EVENTS MUSIC PUBLISHING NEW VENTURES NEWSPAPER RADIO NETWORK
4. 31.4MM unique visitors/month* * source: Ibope, 04/13 globo.com
5. Linked Data at globo.com
6. Semantic Web team
7. Organize and distribute all content produced by Organizações Globo Mission
8. Ontologies
9. Base Current scenario news sports gossip tv
10. Upper Person Place Organization Music Politics Programme Education Sports Future scenario
11. Annotation tool
12. Technologies
13. Embedded into our existing CMSs Web CMS Video publishing system (developed in-house) Common UX for content producers Interface adapts itself to ontology Annotations stored in Virtuoso triple store
15. Interface follows the ontology Fields Search ranges Suggest as you type Automatic entity extraction
16. Automatic page generation
17. globo esporte .com
18. Brainiak: Linked data RESTful API
19. triple store Legacy architecture process queue suggest annotation entity extractor search engine A P I suggest annotation entity extractor entity extractor suggest annotation annotation suggest entity extractor CMS CDA
20. New architecture triple store search engine B R A I N I A K A P I
21. Linked data made simple! Goal
22. Authorization and authentication Reduce need of writing SPARQL queries Single point of access to the triplestore Requirements Data management quality (deduplication, validation, integration with external datasets, etc)
23. RESTful hypermedia API Requirements Performance enhancement - caching
24. Technologies
25. Main concepts
26. Context Isolated dataspace (graph) Sports Context
27. Schema Definition of a data type (i.e. definition of a class) Sports Team Context Schema
28. Sports Team Schema Instance Collection Barcelona Collection Team Context Instance
29. Hypermedia API
30. Content negotiation Decoupling server and client side Access Simplicity (don’t remember URLs, navigate) Hypermedia Application is treated as a state machine
31. self /sports/Team/ Barcelona /sports/ Team inCollection item create delete replace Relations
32. Services
33. Context: List contexts Collections: List collections (within a context) Services Schemas Retrieve schema of a class (collection)
34. List instances (same type) Retrieve instance Instances Create instance Update instance Delete instance
35. { "@context": { "rdfs": "http://www.w3.org/2000/01/rdf-schema#" "sports": "http://semantica.globo.com/sports/", "upper": "http://semantica.globo.com/upper/", }, "@id": "http://semantica.globo.com/sports/Team/Barcelona", "@type": "sports:Team", "rdfs:label": "Barcelona", "upper:fullName": "Futbol Club Barcelona", "upper:acronym": "BARCELONA" } /sports/Team/ Barcelona GET
36. /sports/Team?p=rdfs:label&?o=Barcelona Filtering instances SELECT * FROM <http://semantica.globo.com/sports/> WHERE { ?s a sports:Team . ?s rdfs:label “Barcelona” . }
37. Dealing with legacy and external graphs /sports/ &graph_uri=dbpedia:sports &class_uri=dbpedia-ont:Team
38. Brainiak will be open sourced next Monday at 1st Globo Semantic Day Join now for free: bit.ly/semantic_day_globo Follow us on github: github.com/globocom
39. How we see the future
40. Inference-based navigation SEO (automatic schema.org) Richer content (e.g. timelines), frequent automatic updates Better annotation suggestion (DBpedia Spotlight) Linked with open data (DBPedia, dados.gov.br)
41. THANK YOU Semantic Team semantica@corp.globo.com Ícaro Medeiros globo.com