Towards an Interlinked Semantic Wiki Farm

Slide 1: Towards an Interlinked Semantic Wiki Farm SemWiki2008 - ESWC June 2, 2008 Alexandre Passant1,2, Philippe Laublet1 1 LaLIC, Université Paris-Sorbonne 2 EDF, Recherche & Développement 1

Slide 2: Social software at EDF R&D • Electricité de France R&D • The largest electricity company in France • More than 2000 researchers in its R&D department • Lots of different people: chemists, trading experts, computer scientists ... • Informal and closed-world communities • “Knowledge = Power”, due to its cultural history • Enterprise 2.0 and EDF R&D • Enterprise 2.0 is the use of emergent social software platforms within companies, or between companies and their partners or customers”, Andrew McAfee, May 2006 • Find ways to let people more easily exchange information and build knowledge collaboratively • Introducing new tools and principles • Blogs, RSS feeds, tagging, wikis 2 • Top-down vs bottom-up approach

Slide 3: Using Wikis at EDF R&D • Users adopted wikis for their particular needs, as people do on the Web • Internal project management • Corporate information • Knowledge bases on scientific topics • ... • Usage statistics • More than 1000 registered users for the whole platform • About 50 wikis, 2800 pages, 60 active users • Different wikis as some people want “their” wiki • Closed, read-only and open wikis, depending on the communities • 10% radio between consumers and producers (blogs included) • Many people were not used to those Web 2.0 principles • Both technically and regarding the cultural changes it implies regarding knowledge management 3

Slide 4: Technical limits of wikis for knowledge management • Wikis search-engines can only answer plain-text queries • Natural Language Processing algorithms must be applied to extract information from current wiki systems • Cannot answer questions about the content of wiki pages • “Is EDF located in France ?” • “List all companies known in that wiki” • “Who’s working on tidal energies ?” • Wikis manage documents, not machine-readable representations of real-world objects • Documents and hyperlinks instead of resources, relationships and properties • A gap between these two ways of modeling knowledge • The Semantic Web bridges this gap and semantic wikis provide ways to achieve this 4

Slide 5: From documents to resources 5

Slide 6: Existing semantic wikis • Using wikis to model data using Semanitic Web principles • Ontology population and instances evolution • Adding RDF into wiki pages • SemperWiki • Extending wiki syntax to define annotations • Semantic MediaWiki • Assisting users with user-friendly interfaces • IkeWiki, OntoWiki • Using Semantic Web principles to enrich the usage of wikis • Organizing tags to enhance information retrieval • SweetWiki • Powerful (querying, reasoning, enhanced navigation ...), but raise usability issues in a corporate context • URIs, namespaces, triples... 6 • People need something that work without aditional efforts

Slide 7: UfoWiki: Unifying Forms and Ontologies in a Wiki • A semantic wiki-farm server • Goals • Provide a user-friendly interface to let users create ontology instances and manage them in the wiki way • Hidden semantics for end-users, using forms • Use SIOC and MOAT to model wiki pages meta-data, so that it can be integrated with other internal SIOC data • A corporate SIOC-o-sphere • Connect the meta-data layer to the data (i.e. content) layer • Who wrote facts about EDF ? • Reuse RDF data available on the Web • Geonames.org, DBpedia ... • Interlink data from various wiki instances • While some wikis are private, their data is valuable • Re-use semantic annotations to provide value added interfaces 7 • Macros, semantic search-engine

Slide 8: A form-based user interface • Using forms to maintain ontology instances • Let end-users focus on the content rather than on the modeling • Avoid semantic heterogeneity • Wiki administrators define form-based page templates • Based on existing Drupal modules • Flexinode (Drupal4), rewriting to CCK • Each page corresponds to an ontology class • Organization page => foaf:Organization • Each field corresponds to a property or relationship • Some complex fields can be used to define internal instances • Some fields can be used to define MOAT tags • Ease the process of linking tagged content to related instances • E.g. acronym, nickname • SPARQL autocompletion based on expected class type 8 • Closing the open-world assumption, inference might come later

Slide 9: Using wiki to manage ontology instances • Each created page yields to one main instance, and related / internal ones Macro interne Champ avec auto-complétion Instance interne 9

Slide 10: Linking data and meta-data • embedsKnowledge: linking a sioc:Item to a graph containing the triples that were create from this sioc:Item • Using SIOC Types module for meta-data modeling • sioct:WikiArticle rdfs:subClassOf sioc:Item Meta-data RDF file http:://athena/alex sioc:has_creator EDF Data RDF file embedsKnowledge dc:title Wiki page A athena:EDF rdf:type http://sws.geonames.org/ geonames:locatedIn 3017382 sioct:WikiArticle 10

Slide 11: Architecture of a single wiki edit User 2 Wiki page B Wiki page HTML hyperlink edit A User 1 Document layer (wiki level) produces produces produces produces RDF Store RDF RDF meta-data RDF about page RDF description description meta-data of objects Semantic B of objects about page embedded relationships embedded A in page A between objects in page B uses semantic link semantic link Storage Semantic Web layer uses Meta-data ontologies Data-modeling (SIOC, DC ...) ontologies (SKOS, Domain ontologies ...) 11

Slide 12: Interlinking data from various wikis • All wikis share a common knowledge base • URI identification across wikis • Merge statements about URIs but keep source using named graphs embedsKnowledge embedsKnowledge Wiki page A athena:EDF athena:EDF Wiki page B rdf:type http://sws.geonames.org/ athena:produces 3017382 geonames:locatedIn rdf:type sioct:WikiArticle athena:NuclearEnergy sioct:WikiArticle Wiki A Wiki B stores stores RDF Backend merges athena:EDF geonames:locatedIn athena:produces http://sws.geonames.org/ 3017382 athena:NuclearEnergy 12

Slide 13: Using produced RDF data • RDF data is exported to the triple-store when page is created • Immediately re-usable and up-to-date data • Inline macros • Defined by wiki administrators, using PHP and SPARQL • User-friendly syntax to let end-users embed it in wiki pages • Eg: [onto|members], [onto|type,foaf:Person] • Can be used to run complex queries about data from another wiki • Eg: All activities of an company and related organizations • Direct RDF querying • Advanced users - User-friendly SPARQL interface planned • Queries regarding data, meta-data or both • Semantic search • From keyword to concept • Integration with other SIOC data 13

Slide 14: Macro results • SPARQL query results in wiki pages • Subject of the query is the currently browsed instance • Similar to SemanticMediaWiki inline macros • Semantic back-links 14

Slide 15: SPARQL-ing RDF data from the wiki • Combining meta-data and content levels: • All pages from the wiki #6 that provide information about EDF and than have at least 2 replies select ?page ?title where { graph ?data { :EDF ?predicate ?object } . ?page :embedsKnowledge ?data ; rdf:type sioct:WikiArticle ; dc:title ?title ; sioc:has_container <http://example.org/wiki/6> ; sioc:num_replies ?replies . <http://example.org/wiki/6> a sioct:Wiki . } FILTER (?replies > 1) 15

Slide 16: Reusing RDF data available on the Web • The Linking Open Data initiative • Lots or RDF data available from reference data-sets • GeoNames, DBpedia, riese … • Using the same ontologies in a corporate environment • Low-cost integration • No need to align vocabularies and define mapping between them • GeoNames wrapper • “city, (state), country” fields mapped to geonames.org web-service • Retrieve the location URI and its related RDF file (with coordinates) • Simple way to create geolocation services and enhance navigation • Provide interlinked RDF data 16

Slide 17: Easy-geolocation with our GeoNames wrapper • Combining geolocation and macro-queries • Location of any member of the currently browsed organization • [onto|mapmembers] => SPARQL + rendering 17

Slide 18: Querying the internal SIOC-o-sphere • Combining SIOC-based information from various data sources • Find relevant resource from given keyword with MOAT • Retrieves main / related wiki pages from different wikis • Retrieve “tagged” blog posts 18

Slide 19: Conclusion • Overview of our approach • User-friendly interface with forms mapped to ontologies • SIOC-based meta-data to ease integration with existing SIOC data • Combining meta-data and data (content) layer • Interlinking data from various wiki instances • Using existing RDF data and vocabularies for value-added service • What’s next ? • Validate forms using the underlying ontologies • User-interface to define macros • Linking / reusing more LOD data • Use SIOC in other wikis as a meta-data model 19

Slide 20: Thank you ! Any questions ? slides @ http://apassant.net 20