Semantic Web Implementation Experiences with Annotea and Amaya

Jose Kahan W3C

Introduction

• W3C Semantic Web Advanced Development (SWAD) project
• Provides a framework for rich communication about Web pages through shared RDF metadata
• RDF metadata can be stored locally or in one or more user selected RDF servers
• Two applications:
  • shared annotations

    shared bookmarks

• An open source RDF store

Annotations: introduction

• An annotation is a statement about something

Annotations: basic architecture

Annotations: RDF model

Annotations: an Annotation Instance

<?xml version="1.0" ?>
<r:RDF xmlns:r="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:a="http://www.w3.org/2000/10/annotation-ns#"
xmlns:http="http://www.w3.org/1999/xx/http#"
xmlns:d="http://purl.org/dc/elements/1.1/">
<r:Description>
<r:type resource="http://www.w3.org/2000/10/annotation-ns#Annotation" />
<r:type resource="http://www.w3.org/2000/10/annotationType#Comment" />
<a:annotates r:resource="http://example.org/XDoc.html" />
<a:context>#xpointer(/html[1]/body[1]/ul[5]/li[4]/a[1])</a:context>
<d:creator>Ralph</d:creator>
<a:created>2000-01-10T17:20Z</a:created>
<d:date>2000-01-10T17:20Z</d:date>
<a:body r:resource="file:///home/swick/.amaya/annotations/postit.html" />
</r:Description>
</r:RDF>
<?xml version="1.0" ?>

Annotations: annotation reply RDF model

Annotations: interacting with the Annotea RDF store

If time, go thru the protocol document, otherwise, use the following content

• Standard HTTP methods for publishing (PUT), updating (PUT, POST), deleteting (DELETE)
• Problem: how to publish an RDF statement without knowing its URI?

  --> solution: have the server assign it an URI and discard the previous one

• Retrieval done thru a customized HTTP GET Query

  GET /Annotation?w3c_annotates=http://serv1.example.com/some/page.html HTTP/1.1
  GET /Annotation?w3c_reply_tree=http://annotea.example.org/Annotation/3ACF6D754 HTTP/1.1

  --> works, but this is not a generic RDF store query (see Algae)

• Retrieval of annotations metadata is separate from the retrieval of the annotations body

  --> download the body only if you need to see it

Annotations: interacting with the Annotea RDF store (cont)

• Algae: Specific RDF store query language (predates the DAWG WG)
• Example: query for getting annotations with their properties and values:

(namespace (a http://www.w3.org/2000/10/annotation-ns#)

ask '(
    (a::annotates ?annot http://example.com/page.html)
    (?property ?annot ?value))
:collect (?annot ?property ?value)) 

Annotations: clients

• Amaya <-- supports all the features
• Annozilla (Mozilla)
• Snufkin (IE)
• bookmarklet (javascript, most browsers)

Annotations: current issues

• Annotating HTML documents
  ->XPointers apply only to XML docs.
• Annotating live documents
  -> avoiding misleading annotations
  -> supporting the revision process

Contributions to the field

Problems, !MVC, keeping all the RDF data

Bookmarks: introduction

• Annotea Bookmarks : helping share bookmarks between different browsers and users
• RDF schema
  • Supports all of the XBEL features, while introducing better semantics

    --> Topics instead of folders

    --> Each topic has a URL (simplifies sharing)

  <rdf:Description rdf:about="ambookmark1">

       <bm:hasTopic rdf:resource="amtopic1" / >

  </rdf:Description>
      

• Other features: a bookmark can belong to multiple topics, rdf:seeAlso, multiple views of the same bookmark file

Bookmarks: RDF model for bookmarks

• n.b.: b:lastVisited property is missing in the diagram

Bookmarks: RDF model for topics

Bookmarks: a topic hierarchy

Bookmarks: organizing bookmark items with RDF collections

• Each topic is associated with an RDF collection

 <rdf:Description rdf:about="file:///home/kahan/.amaya/bookmarks.rdf#Topic1">

  <b:collection rdf:parseType="Collection">     <rdf:Description
         rdf:about="file:///home/kahan/.amaya/bookmarks.rdf#bookmark2" /> 
    <b:Separator /> 
   <rdf:Description
         rdf:about="file:///home/kahan/.amaya/bookmarks.rdf#Topic4" /> 
    <rdf:Description 
         rdf:about="file:///home/kahan/.amaya/bookmarks.rdf#bookmark1" />
  </b:collection>

</rdf:Description> 

Bookmarks: exercise

• Question: how can we assing a dc:title to a seealso?

Bookmarks: exercise (cont)

• Answer: use reification: a blank node plus a dc:title

Bookmarks: new possibilities

• Associating a URL with a bookmark file
• Opening more than one bookmark file
• Publishing bookmark files to a Web server using HTTP PUT, WebDAV, scp, ...

Bookmarks: new possiblities (cont)

Bookmarks: new possibilities (cont)

• Possiblity to separate the user browsing properties from the other bookmark properties

  --> Different user "browsing preferences" for the same bookmark file

  --> Possible to have different views of a same file using this principle and collections

Bookmarks: relocation of resources problem

• Bookmarks can be stored locally in files or in a Web server
• Refering to a resources inside a file when the file moves around the Web

  <rdf:Description rdf:about="amtopic1">

       <dc:creator>Jose</dc:creator>

  </rdf:Description>

• Relative URLs: change when the file changes location
• URNs: hard to resolve
• When publishing statements to an RDF datastore, URIs can be changed

Bookmarks: chosen solution: URNs + owl:sameAs

• Assign a URN to each resource in a local file

  <rdf:Description rdf:about="urn:uid:123-4325-53">

       ...
  </rdf:Description>

• Moving the file around doesn't break references to the resource
• However, the client needs to know where a given URN is found

  --> keep the resource in the same bookmkark file, assume the clients knows where are those files

• Whenver the URI of the resource changes, record its previous URIs using owl:sameAs

  <rdf:Description rdf:about="http://annotea.w3.org/ab213450">
   <owl:sameAs rdf:resource="urn:uid:123-4325-53" / >
       ...

  </rdf:Description>
      

• Solution depends on robustnees of URN.

Bookmarks: conclusions

• RDF is most appropriate for this kind of application
• RDF part of the application was done very quickly:

  --> UI represented most of the work

  --> adding new features is very easy (multiple topics and importing a topic hierarchy, a couple of hours)

• Possible to process and display the bookmark file with other tools

Bookmarks: future work

• Take into account the SWADE Thesaurus activity

  --> use the thesaurus paradigm to define bookmark topics

  --> create (or populate) thesaurus while making bookmarks

• Finish integrating the URN solution (code written, going thru discussions)
• libannotea: move our development from Amaya to Mozilla (see below)

Developing in the Semantic Web: parsing

• Raptor (used in Amaya/annotations)
  • RDF parser developed by Dave Becket
  • Callback programming model: a callback each time a triple is found
  • Problems: having to construct an RDF model in memory by hand
  • Problems: loosing RDF statements you don't know how to handle
  • Possible area of interest: quick parsing (read only, but not updates)

Developing in the Semantic Web: parsing (cont)

• Redland (used in Amaya/bookmarks)
  • RDF Application framework (similar to DOM, but for RDF)
  • Builds an RDF model in memory
  • Has an API for querying the model
  • Can serialize the model as a flat file or in a database
  • Possible areas of interest: interacting with an RDF model (parsing, editing)
  • Benefits: application code only makes queries or adds statements

    --> small code that makes complex things

    --> programming using an MVC paradigm

LibAnnotea

• Moving our development from Amaya to Mozilla

  --> better visiblity

Conclusions: my experience

• Learn to expect the unexpected

  --> RDF data missing properties or having more than one property... it's valid!

• Take into account reification

  --> rdfs:seeAlso and dc:title

• Take into account containers (collections, seq, ... ) as much as possible, ignore if not used

  --> Could have been used for rdf:seeAlso in bookmarks

• RDF can be serialized in any way and still be legal

  --> Don't assume an XML structure order (as RSS does)

• RDF allows to combine different properties and aggregate more as needed

  --> if it breaks, it probably has a modelizing problem

• Moving RDF stores or files can be complex

More Information

• http://www.w3.org/2001/Annotea/
• http://www.w3.org/Amaya/ (annotation client)
• http://annotest.w3.org/ (annotation server)
• www-annotations@w3.org (archive)
• public-annotea-dev@w3.org (archive)
• http://www.w3.org/
• http://www.redland.opensource.ac.uk/