Doing Clever Things with the Semantic Web


Published on

Invited talk at the EPIA 2011 conference

Published in: Technology, Education
1 Like
  • Be the first to comment

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • Would be nice to have a bit of loted here…
  • Doing Clever Things with the Semantic Web

    1. 1. Doing Clever Things With the Semantic Web<br />Mathieu d’Aquin<br />Knowledge Media Institute, the Open University<br />
    2. 2. The Semantic Web<br />Using the Web to publish, share and exploit information/knowledge<br />From machines to machines<br />Using graph-based data modeling, knowledge representation (ontologies) and reasoning<br />
    3. 3. Linked Data<br /><br />As set of principles and technologies for a Web of Data<br />Putting the “raw” data online in a standard representation (RDF)<br />Make the data Web addressable (URIs)<br />Link to other Data <br /><br />
    4. 4. Galen<br />NCI<br />…<br />Music<br />DC<br />WORDNET<br />RSS<br />TAP<br />FOAF<br />…<br />…<br />…<br />…<br />Metadata<br /><rdf:RDF><br /><channel rdf:about=“”><br /><title>Elementaries - The Watson Blog</title><br /><link></link><br /><description><br />"Oh dear! Where the Semantic Web is going to go now?" -- imaginary user 23<br /></description><br /><language>en</language><br /><copyright>Watson team</copyright><br /><lastBuildDate>Thu, 01 Mar 2007 13:49:52 GMT</lastBuildDate><br /><generator>Pebble (</generator><br /><docs></docs><br />…<br /><rdf:RDF><br /> <foaf:Imagerdf:about=''><br /> <dc:title>Zen wisteria</dc:title><br /> <dc:description></dc:description><br /> <foaf:pagerdf:resource=''/><br /> <foaf:topicrdf:resource=''/><br /> <foaf:topicrdf:resource=''/><br /> <dc:creator><br /> <foaf:Person><foaf:name>Mathieu d'Aquin</foaf:name><br />…<br /><rdf:RDF><br /> <owl:Ontology rdf:about=""><br /> <owl:imports rdf:resource=""/><br /> </owl:Ontology><br /> <j.1:Organization rdf:ID="KMi"><br /> <rdfs:comment rdf:datatype=""<br /> >The Knoledge Media Institute of the Open University, Milton Keynes UK</rdfs:comment><br /> </j.1:Organization><br /> <j.1:Document rdf:ID="KMiWebSite"><br /> …<br />UoD<br />
    5. 5. The Semantic Web<br />Knowledge/<br />Problem Solving Methods<br />Semantic Web Applications<br />Doing Clever Things With the Semantic Web<br />Intelligent Agent<br />Smart Features<br />Clever Things…<br />
    6. 6. What I Want to Talk About<br />Using the Semantic Web as A Knowledge Base<br />KMi Watson and Finding Ontologies<br />Doing More with Links<br />Exploring the Web of Data<br />Back to the Future<br />AI + Linked Data = Semantic Web? <br />
    7. 7. Using the Semantic Web<br />Need for a Gateway to the Semantic Web<br />Dynamically retrieving, exploiting and combining relevant semantic resources from the Semantic Web<br />
    8. 8. KMi Watson<br />
    9. 9. Architecture<br />
    10. 10. Interface<br /><br />Watson: More than a Semantic Web Search Engine, Semantic Web Journal<br />
    11. 11. Watson as a Service<br />Providing Web accessible APIs to a collection of online ontologies and semantic data sources<br />
    12. 12. Chose an entity to search<br />Integrate statements <br />Into the edited ontology<br />Get entities from online ontologies<br />Example Application: Ontology Construction<br />Reusing Knowledge from the Semanrtic Web with the Watson Plugin, Demo at ISWC 2008<br />
    13. 13. Concept Relation Discovery<br />SeaFood<br />Meat<br />wine.owl<br />AcademicStaff<br />Semantic Web<br />Semantic Web<br />Researcher<br />ka2.rdf<br />Meat<br />SeaFood<br />Ham<br />pizza-to-go<br />NALT<br />AcademicStaff<br />Researcher<br />Ham<br />SeaFood<br />ISWC<br />SWRC<br />NALT<br />Agrovoc<br />
    14. 14. Exploring the Semantic Web as Background Knowledge for Ontology Matching, Journal of Data Semantics<br />
    15. 15. PowerAqua: Question Answering<br />
    16. 16. From Semantic Web Research to Linked Data Applications<br />Watson as a platform to research applications and techniques on top of semantic web resources<br />But how can the Semantic Web be exploited and used in real-world application?<br />Starting from what we know best…<br />
    17. 17. Applying Linked Data<br />The Open University is the largest University in the UK, where all the courses are realized at a distance<br />Creating the first University Linked Data platform:<br />Demonstrate the value of the technology and push the research through real-world scenarios<br />
    18. 18.<br />
    19. 19. Applications<br />Mobile and Personal Semantics<br />Social<br />Resource Discovery<br />Research<br />Exploration<br />
    20. 20. Example application: Finding relevant resources<br />Zablithet al, LinkedLearning 2011<br />
    21. 21. Data as Web resources, accessible everywhere<br />
    22. 22. See also: Zablith et al., COLD 2011<br />
    23. 23.<br /><br />
    24. 24. Supporting Researchers: The Reading Experience Database<br /><br />40,000 accounts of somebody reading something at some time in some place<br />Used by researchers in literature and history to explore research hypotheses<br />
    25. 25. Event<br />Location<br />locatedIn<br />subClassOf<br />subClassOf<br />Experience<br />City<br />Country<br />date: Date<br />readerInvolved<br />originCountry<br />textInvolved<br />occupation<br />givesBackgroundTo<br />Person<br />religion<br />gender<br />creator/editor<br />LinkedEvent Ontology<br />Document<br />CITO Citation Ontology<br />Dublin Core<br />title: String<br />description: String<br />published: Date<br />providesExcerptFor<br />FOAF<br />DBPedia<br />
    26. 26.
    27. 27. Back to the Future<br />The Semantic Web is both a vision and a reality<br />Making the Web more than a network of documents: the biggest, most distributed knowledge base ever<br />What could AI do with such a knowledge base?<br />
    28. 28. Linked Data Mining<br />Finding unexpected patterns in the use of the distributed data graph<br />
    29. 29. Linked Data Mining: Example<br />Using Formal Concept Analysis + Reasoning to build a hierarchy of questions a linked dataset can answer<br />Use statistical metrics to identify the ones that are most likely to be interesting<br />Extracting Relevant Questions to an RDF Dataset Using Formal Concept Amalysis at KCAP 2011<br /><br />
    30. 30.
    31. 31.
    32. 32. Reasoning<br />To analyze and understand raw data in relation with online resources<br />Example: Online personal information management<br />Online Activities Ontology <br />HTTP Ontology <br />Parameters and Website info.<br />Web Site Information<br />Personal Information<br />Trust Model<br />Location Information<br />
    33. 33. Enriched with linked data<br />Google Services<br />Entertainment Websites<br />Web Analytics<br />Internet Search Engine<br />subject/category<br />Video sharing<br />Video Hosting<br /><br />Company<br />developer<br />Web Search Engine<br />Search Engine<br />type<br />subject/category<br />google<br />owner<br />subsediaryOf<br /><br /><br />parent<br />DBpedia<br />freebase<br />
    34. 34. Basic processing/analysis<br />Requests by time of day<br />Requests by User Agents<br />Interests<br />Trust<br />
    35. 35.<br /><br />
    36. 36. Understanding knowledge representation and data modeling<br />The Semantic Web also represents a very large, collaborative base of formally represented knowledge <br />This can also be mined, to discover things about knowledge representation and data modeling<br />
    37. 37. Ontologies on the Semantic Web<br />Underlying description logic<br />Number of entities<br />Domain covered<br />
    38. 38. Relationships between ontologies<br />DOOR: Towards a Formalization of Ontology Relations at KEOD 2009<br />
    39. 39. Detecting versions of ontologies<br />When published on the Web, the information about the evolution of ontologies is lost<br />Using URI patterns to find candidate versions of ontologies<br /> <br /> <br />Applying machine learning algorithms (SVM, Naïve Bayes and Decision tree to recognize chains of versions of ontologies<br />Allocca at ESWC 2011<br />Obtained 90% Precision (SVM)<br />Collected thousands of ontology version sequences to be analysed<br />For example, distribution of similarity in version and non-version ontologies (right)<br />
    40. 40. Agreement/Disagreement between ontologies<br />Ontologies are knowledge artifacts, they express opinions and beliefs and contradict each others<br />Assessing (dis)agreement in ontologies is very useful to understand how to combine knowledge from different sources<br />
    41. 41. Assessing Statements related to SeaFood<br />Nb1: #ontologies in which the statement appears.Nb2: #ontologies containing entities matching the subject and object of the statement. <br />a: global agreement, d: global disagreement, cs: consensus, ct: controversy<br />
    42. 42. 21 different ontologies with a SeaFood concept<br />Disagreement<br />Agreement<br /> Formally Measuring Agreement and Disagreement in Ontologies at K-CAP 2009<br />
    43. 43.<br />Vegan subClass Vegetarian<br />SeaFoodsubClassOf Meat<br />SeaFooddisjointWith Meat<br />
    44. 44. The brighter the blue the higher the positive consensus (higher agreement)<br />The brighter the red the lower the negative consensus (higher disagreement)<br />Dark = controversy: no clear cut between disagreement and agreement<br />Example: The statements attached to the class Employee are controversial: some ontologies agree, others disagree (often due to alternative representations of roles)<br />AKT Portal<br />Using consensus to assess an ontology(a new NeOn toolkit plugin<br />Visualizing Consensus with Online Ontologies to Support Quality in Ontology Development at ONTOQUAL@EKAW 2010<br />
    45. 45. So my point is…<br />The Semantic Web is a fantastic open field for AI<br />It is going to become omnipresent, hidden, personal<br />Exploring, Exploiting and Excavating the Semantic Web for <br />Research in technology (creating it and studying it) <br />Research in other areas<br />Everyday tasks <br />Still, after 10 years of research, represent new directions for many fields of the AI community, with their own issues, challenges and applications<br />
    46. 46. Thank You!<br />More at:<br /><br /><br />@mdaquin<br />
    1. A particular slide catching your eye?

      Clipping is a handy way to collect important slides you want to go back to later.