ISO MLR semantics


Published on

Published in: Education
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

ISO MLR semantics

  1. 1. Mikael Nilsson < [email_address] > et al. Making MLR semantic?
  2. 2. History <ul><li>Contribution in Jeju from a number of experts: </li></ul><ul><ul><li>“Requirements for ISO MLR interoperability”(WG4 N0238) </li></ul></ul><ul><ul><li>Suggested basing MLR on semantic technologies </li></ul></ul><ul><li>The group was tasked to present a proposal for an MLR standard based on semantic technologies </li></ul><ul><li>Outline of such a proposal was submitted (WG4 N0278) - presented here </li></ul>
  3. 3. Background <ul><li>Islands of metadata interoperability, for example </li></ul><ul><ul><li>The “LOM island” - IEEE LOM and LOM-based profiles </li></ul></ul><ul><ul><li>The “MODS island” </li></ul></ul><ul><ul><li>The “MPEG-7 island” </li></ul></ul><ul><ul><li>The “Dublin Core & RDF island” </li></ul></ul><ul><li>Two approaches to Application Profiles </li></ul><ul><ul><li>Base standard – profiles customize the base </li></ul></ul><ul><ul><li>(LOM, MODS, MPEG-7) </li></ul></ul><ul><ul><li>Framework only, profiles combine terms arbitrarily (Dublin Core, RDF) <-- MLR wants to be here (?) </li></ul></ul>
  4. 4. Metadata interoperability now IEEE LOM NorLOM UK LOM Core ... RDF Dublin Core DC APs Semantic Web ... MARC21 MARC-XML METS METS MPEG-7
  5. 5. Metadata interoperability vision Education Government Libraries Multimedia Semantic Web
  6. 6. Levels of interoperability <ul><li>Human interoperability - words </li></ul><ul><ul><li>Use the same definition of words , regardless of technical framework </li></ul></ul><ul><li>Semantic interoperability – the cloud </li></ul><ul><ul><li>Machines apply the same processing to terms whereever they appear </li></ul></ul><ul><ul><li>This is the purpose of RDF </li></ul></ul><ul><li>Profile interoperability – the domain </li></ul><ul><ul><li>Domain-specific interoperability based on shared profiles, vocabularies, etc. </li></ul></ul><ul><ul><li>Quality control, syntax validation etc. </li></ul></ul>
  7. 7. Proposal for MLR <ul><li>DON'T create a new metadata island </li></ul><ul><ul><li>DON'T create a need for more crosswalks </li></ul></ul><ul><li>DO use a framework-based approach </li></ul><ul><ul><li>DO allow for application profiles combining terms from other sources </li></ul></ul><ul><li>DON'T reinvent the framework </li></ul><ul><ul><li>DON'T require others to redefine their terms for use in MLR </li></ul></ul><ul><li>DO base the framework on the RDF model </li></ul>
  8. 8. Statement-based models
  9. 9. Statements as graphs title contribution date entity name My learning resource Contribution A Person B “ A book” “ John Smith” “ 2008-09-03”
  10. 10. <ul><li>An XML format defined from an application profile </li></ul><ul><li>The format depends on the application profile </li></ul><ul><li>Interpreting as RDF triples is straightforward if application profile is knowns </li></ul>Making an XML schema <LearningResource> <Title>A book</title> <Contribution> <Date>2008-09-03</Date> <Entity> <Name>John Smith</Name> </Entity> </Contribution> </LearningResource>
  11. 11. Another example (FOAF-like) <Person uri=””> <Name>John Smith</Name> <Email> [email_address] </Email> <Knows uri=”” /> </Person> foaf:name foaf:mbox foaf:knows “ John Smith” “ [email_address] ”
  12. 12. Walk-through <LearningResource grddl:transform=”http://yyy/mlr.xsl” > <Title>A book</title> <Contribution> <Date>2008-09-03</Date> <Entity> <Name>John Smith</Name> </Entity> </Contribution> </LearningResource> http://yyy/mlr.xsl foaf:name foaf:mbox foaf:knows title contribution date entity name foaf:mbox foaf:knows “ John Smith” “ [email_address] ” My learning resource Contribution A “ A book” “ John Smith” “ 2008-09-03” “ [email_address] ” My learning resource Contribution A Person B “ A book” “ John Smith” “ 2008-09-03” title contribution date entity name
  13. 13. “Follow your nose” RDF Schema label comment range HTML FOAF specification “ Knows” “ A person known by this person (indicating some level of reciprocated interaction between the parties)” foaf:Person
  14. 14. Linked Open Data <ul><li>More than 2 billion RDF triple </li></ul>
  15. 15. Summary <ul><li>Semantic technologies allow for </li></ul><ul><ul><li>Large-scale interoperability (triples, AP-independent, follow-your-nose, linked data) </li></ul></ul><ul><ul><li>Ontology support </li></ul></ul><ul><ul><li>Reuse of existing standards </li></ul></ul><ul><ul><li>Collaboration between standards bodies </li></ul></ul><ul><ul><li>Reuse of existing tools </li></ul></ul><ul><ul><li>Implementation in many environments </li></ul></ul><ul><ul><ul><li>From mobile or AJAX applications </li></ul></ul></ul><ul><ul><ul><li>Through HTML (RDFa) and RSS </li></ul></ul></ul><ul><ul><ul><li>To multi-billion-triples RDF stores </li></ul></ul></ul>
  16. 16. Consequences for MLR drafts <ul><li>Use a “statement”-based model based on RDF </li></ul><ul><ul><li>Mature specification, large set of tools (parsers, reasoners, databases, etc) </li></ul></ul><ul><ul><li>Ontology support (formal semantics) </li></ul></ul><ul><li>REMOVE structure attributes from MLR terms </li></ul><ul><li>Provide new MLR templates for describing terms </li></ul><ul><ul><li>Properties, like “title”, “creator” </li></ul></ul><ul><ul><li>Classes, like “Learning Resource”, “Event”, “Contribution”, “LangString”, “Classification” etc. </li></ul></ul><ul><li>REMOVE current application profile definition </li></ul>
  17. 17. Defining a property
  18. 18. Roadmap <ul><li>Part 1: Overview of MLR, how to make new parts, etc </li></ul><ul><li>Part “B”: Basic RDF-based model, templates for properties, etc. </li></ul><ul><li>Part “C”: Core elements. Allow reference to e.g. Dublin Core terms. </li></ul><ul><li>Part “D”: Definition of Application Profiles, records, etc. </li></ul><ul><li>Part “E”: MLR Core Application profile </li></ul><ul><li>Part “F”: XML format for MLR application profiles </li></ul>
  19. 19. Some issues <ul><li>Should MLR use RDF directly or add some constructs on top? </li></ul><ul><li>Can properties from other specs be reused in MLR application profiles? (Dublin Core etc.) </li></ul><ul><li>Should MLR parts reference external properties? </li></ul><ul><li>Should MLR collaborate directly with e.g. DCMI in developing notion of application profile? </li></ul>