SweoIG/TaskForces/InfoGathering/DataVocabulary

From W3C Wiki
Jump to: navigation, search

Suggested Vocabulary for your data to be syndicated

For items, we use a combination of RSS, dublin core, DOAP. if you need special vocabs, use them as needed.

If two properties from different vocabularies can be used (for example rss:title or dc:title), we favor the most "popular" one, for example RSS is extremly popular and outrules RDF and RDFS. The order of preference is:

  • RSS
  • RDF
  • RDFS
  • DC
  • DOAP
  • SKOS
  • SY(ndication)

Classes and their proposed properties:

  • rss:channel - a list of resources, a compiled collection of resources
    • Properties:
    • rss:title
    • rss:link
    • rss:description
    • dc:creator
    • dc:date
    • sy:updatePeriod
    • sy:updateFrequency
    • sy:updateBase
    • rss:items (and the usualy rdf:Seq of rss:Item inside)
  • rss:item - an information resource
    • rss:title
    • rss:link
    • rss:description
    • dc:date
    • dc:language
    • dc:rights
    • skos:subject - to point to elements defined in the ../ClassificationOntology

For Software tools:

For books or articles:

  • dc:creator

For people and organisations:

  • foaf:Person (DOAP and many aggregators use it)
  • foaf:Organisation

For ontologies:

  • website-URL VS ontology download/ns-URL

For publications/bibtex:

Example for books:

http://www.bibsonomy.org/swrc/tag/sweo%2Bbooks?items=300

with the real page at:

http://www.bibsonomy.org/tag/sweo%2Bbooks

About the SWEO Data Vocabulary

Used to create lists of resources that are interesting for Semantic Web people.

What we need to capture:

  • all elements listed in the ../ClassificationOntology (article, blog, meeting, book, tutorial, etc...)
  • lists of such elements
  • a rating of elements (1-10, stars, favorites, etc)

suggestion is reusing a mixture of existing, popular vocabularies and a new ontology for classification

  • SKOS: for our concept map of terms we use for classification (book, article, etc) and for suggested keywords/tags (to be compatible with del.icio.us)
  • RDFS: comment
  • RSS 1.0: date, title, uri, text
  • Dublin Core (obvious)
  • DOAP: for projects
  • rev: Reviews
  • DC Types: Specifying doap:Version types (e.g. dctype:Collection, dctype:Dataset, dctype:InteractiveResource, dctype:Service)

Most important decision taken: To be generic and reusable, all items are described using RSS 1.0 and additional vocabularies. All items have an rdf:type of rss:item and properties rss:title and rss:link. Additionally, for people they have the type foaf:Person, for project doap:Project (etc). For types and tags that are not "obvious", they are annotated with SKOS using the item as subject and a skos:subject triple having a concept as object. A taxonomy of concepts of interest, this taxonomy is called ../ClassificationOntology, expressed using SKOS and managed by SWEO and others.

Alternatively to SKOS, we could have also used rdf:type and rdfs:Class to classify things, Ivan, Benjamin and Leo voted in favor of SKOS and common types.

Feedback:

LeoSauermann: I think that using skos allows us to have something like a "tagging" feature on the website, where skos:prefLabel is the name of the tag and each tag is represented as skos:Concept. The relation between an rss:item and such a concept is then item-skos:subject-skos:Concept.

BenjaminNowack prefers proper types (where available) instead of an RSS-only approach, as the portal is probably going to describe non-document resources which may not really fit under the feed item umbrella. For the tagging stuff: SKOS. IvanHerman agrees with that.... LeoSauermann agrees.....PasqualePopolizio agrees.

This page is based on

To generate this vocabulary and howto, these sources have been consulted

If we want to add this, the list of W3C Team and WG members' presentations on SW is always available in RDF, too, either by this URI or by accessing and SPARQL-ing the data sources directly.