Linkator: enriching web pages by automatically adding dereferenceable semantic annotations
Download as PPTX, PDF2 likes993 views
The document presents Linkator, a tool for enhancing web pages through automatic semantic annotations, enabling better navigation and linking of content. It outlines the challenges of automatic linking, including term identification, disambiguation, and target resource selection, and discusses the Linkator's approach to solve these problems using an information extraction engine and semantic links. The authors emphasize the potential for future improvements and broader applications of Linkator in various domains.
![State-of-the-Art in Automatic LinkingCandidate Terms:Focused on term disambiguation using an auxiliary knowledge base or dictionaries (e.g. wikipedia and wordnet).Link Target:It is selected from a specific knowledge base [1] or from a collection [2] of target documents. LimitationsDoes not support well users interested in a broader range of domains.[1] Mihalcea, R. and Csomai, A. Wikify!: linking documents to encyclopedic knowledge. In Proceedings of the 16th ACM Conference on Information and Knowledge management (CIKM 07), Lisbon, Portugal, pp. 233-242, 2007.[2] Gardner JJ, Krowne A, Xiong L. NNexus: An Automatic Linker for Collaborative Web-Based Corpora. IEEE Trans. Knowl. Data Eng. 21(6). 829-839. 2009.](https://image.slidesharecdn.com/linkator-presentation-v5-101210102135-phpapp02/85/Linkator-enriching-web-pages-by-automatically-adding-dereferenceable-semantic-annotations-6-320.jpg)
![State-of-the-Art in Automatic LinkingExample: Wikify! [1] is focused on linking keywords on web pages to Wikipedia articlesNnexus [2] focus on linking keywords obtained from an index extracted from target documents. [1] Mihalcea, R. and Csomai, A. Wikify!: linking documents to encyclopedic knowledge. In Proceedings of the 16th ACM Conference on Information and Knowledge management (CIKM 07), Lisbon, Portugal, pp. 233-242, 2007.[2] Gardner JJ, Krowne A, Xiong L. NNexus: An Automatic Linker for Collaborative Web-Based Corpora. IEEE Trans. Knowl. Data Eng. 21(6). 829-839. 2009.](https://image.slidesharecdn.com/linkator-presentation-v5-101210102135-phpapp02/85/Linkator-enriching-web-pages-by-automatically-adding-dereferenceable-semantic-annotations-28-320.jpg)
Linkator: enriching web pages by automatically adding dereferenceable semantic annotations
- 1. Linkator: enriching web pages by automatically adding dereferenceable semantic annotationsSamur Araujo, Geert-Jan Houben, Daniel SchwabeWeb Information SystemsDelft University of Technology, the Netherlands
- 2. Summary – dereferencing semantic annotationsWhat dereferencing semantic annotations is about?Automatic linking web pages.SummaryOverview of the problem and motivation.Our approach for solving the problem.One example of use.
- 3. MotivationLinks between HTML pages are the main mechanism to navigate on web pages.However, a lot of pages are unlinked or poorly linked.Terms on pages have meaning and are intrinsically associated to concepts or entities that the user is interested in.These terms can be interpreted by machines and automatically linked to relevant resources on the web.
- 5. Problem Statement The problem of automatic linking can be divided in 3 sub-problems:How to identify candidate terms (anchors) for adding links?It denotes concepts in which the user is interested. Which concept does a candidate term represent?Disambiguate a candidate term.How to identify a web resource to be the link target?How to select a source of data for finding the destination of the link?
- 6. State-of-the-Art in Automatic LinkingCandidate Terms:Focused on term disambiguation using an auxiliary knowledge base or dictionaries (e.g. wikipedia and wordnet).Link Target:It is selected from a specific knowledge base [1] or from a collection [2] of target documents. LimitationsDoes not support well users interested in a broader range of domains.[1] Mihalcea, R. and Csomai, A. Wikify!: linking documents to encyclopedic knowledge. In Proceedings of the 16th ACM Conference on Information and Knowledge management (CIKM 07), Lisbon, Portugal, pp. 233-242, 2007.[2] Gardner JJ, Krowne A, Xiong L. NNexus: An Automatic Linker for Collaborative Web-Based Corpora. IEEE Trans. Knowl. Data Eng. 21(6). 829-839. 2009.
- 7. Linkator Approach LinkatorExtract Terms from Web PagesAssociate Terms to ConceptsFind Resources that Represents these ConceptsCore LinkatorInformation Extraction EngineSemantic Annotator
- 8. Link ClickedPage AccessedPage is accessedAnnotated pageTerm are extractedAnnotation is extractedPage is semantically annotatedEndpoint is chosenSemantic Links createdQuery is formulatedIf notfoundSearch for a resource
- 9. Linkator Approach Web BrowserLinkator Client - Firefox PluginAnnotatorRDFa AnnotatorInformation Extraction EngineHTTPHTTPLinkator ServerLinked DataEndpoint ResolutionSparqlQuery Formulation
- 10. Semantic Link – Definition A semantic link is an HTML tag A that is semantically annotated with RDFa.It contains RDF triples associated to it.Semantic Link causes a query over Linked Data.
- 11. RDF Triples associated to the Semantic LinkSemantic Links
- 12. Dereferencing Semantic LinksLinkator uses the Linked Data cloudfor discovering a destination for the semantic link as opposed to querying search engines or a fixed knowledge base.Algorithm for Endpoint ResolutionAlgorithm for Query Formulation
- 13. Endpoint ResolutionTask:Find endpoints that contain a specific concept.Linkatorselects available endpoints based on the vocabulariesused in the semantic links. voiD(Vocabulary of Interlinked Datasets)
- 14. Endpoint ResolutionSelect the vocabulary of all RDF types associated with the annotation.Or select the vocabularies of all predicates associated with the annotation.
- 15. Endpoint ResolutionThe SelectEndpoint function find the resource: http://ontoware.org/swrc/swrc_v0.3.owl#AuthorIt extracts the vocabulary associated with this resource:http://ontoware.org/swrc/swrc_v0.3.owlIt queries the voiDdescriptor of the available SPARQL endpoints, looking for such a vocabulary.
- 16. Query FormulationQuery is based on the object of the triple.Try to find a human-readable representation of the resource, i.e., try to match predicates such as: foaf:homepage, akt:has-web-address, rdfs:seeAlso.
- 17. Proof of ConceptSemantic links for pages that contain bibliographic citations. Extended version of FreeCite parsing engine.Example of bibliographic citation:Keesvan derSluijs, Geert-Jan Houben, Erwin Leonardi, Jan Hidders. Hera: Engineering Web Applications Using Semantic Web-Based Models. Book chapter: Semantic Web Information Management: A Model-Based Perspective, De Virgilio, Roberto; Giunchiglia, Fausto; Tanca, Letizia (Eds.), Chapter 22, 2010, Springer.
- 18. LinkatorExtract Terms from Web PagesAssociate Terms to ConceptsFind Resources that Represents these ConceptsCore LinkatorInformation Extraction EngineSemantic AnnotatorHtml PageSparql Endpoint Discovering and SelectionMarkupRemovedEntity Extraction Plain TextText Semantically AnnotatedEndpoint QueryingSemantic link clickedSemantic AnnotationInsert annotations on the pageHTML Page Semantically AnnotatedURL GenerationFreeCite Extraction EngineCore Linkator
- 19. Example – HTML Page without Links
- 21. Example – Page annotated with RDFa
- 22. Example – Page with Semantic Links
- 25. Conclusion and Future WorkFor a specific scenario of linking bibliographic citations Linkator provides a reasonable solution. The composition of the Semantic Web technologies can provide a reasonable solution for the problem of automatic linking.Linkator is a concrete application that uses Semantic Web technologies.Future Work: Use Linkator in a broader scenario.Enhance the Linkator algorithms.Evaluate the precision and recall of the linking.
- 26. Questions?Thank you for your attention!Samur Araujos.f.cardosodearaujo@tudelft.nlYou can download Linkator at:http://www.wis.ewi.tudelft.nl/
- 27. Annotation on the page are used to find the link destinationAnnotated HTML PageHTML PagePage is annotatedLink is clickedRDF
- 28. State-of-the-Art in Automatic LinkingExample: Wikify! [1] is focused on linking keywords on web pages to Wikipedia articlesNnexus [2] focus on linking keywords obtained from an index extracted from target documents. [1] Mihalcea, R. and Csomai, A. Wikify!: linking documents to encyclopedic knowledge. In Proceedings of the 16th ACM Conference on Information and Knowledge management (CIKM 07), Lisbon, Portugal, pp. 233-242, 2007.[2] Gardner JJ, Krowne A, Xiong L. NNexus: An Automatic Linker for Collaborative Web-Based Corpora. IEEE Trans. Knowl. Data Eng. 21(6). 829-839. 2009.
- 29. Endpoint ResolutionFUNCTION SelectEndpoint E := Array R : = select all rdf:type objects associated to the semantic link T := ExtractVocabulary(R)FOR EACH vocabulary in T DO{E.add (select endpoints that contain this vocabulary) }IF E = Empty { R := select all predicates associated to the semantic link T := ExtractVocabulary(R)FOR EACH vocabulary in T DO {E.add (select endpoints that contain this vocabulary) } }RETURN E FUNCTION ExtractVocabulary(R) V := ArrayFOR EACH resource in R DO {V.add (extract the vocabulary from the resource) }RETURN V12345678910111213141516171819202122232425262728
- 30. Semantic Link – ExampleTriples associated with the semantic link.
Editor's Notes
- #2: I am in the start phase of the phd research. In this presentation, I will outline the vision at the start of the phd period on the research problem which is building trust in web content and our approach to solving this problem. Also I will give a brief plan of my PhD research.
- #6: We focus on content trust and formulate our main research questions. The first key issue here is to investigate what kind factors that can influence trust in content.Following the first one, we also need to know how to capture and represent the information about these factors.The third key issue is how to assess or compute content trust based on the information we get from the second step. Ideally we want to have a trust value assigned to every piece of content. Different from the propagation of trust through the network of people, since we now have more information, and semantics about the content, we want to build metrics to assess the trustworthiness based content itself and the connection between different pieces of content, especially the semantic similarity and relation.