MIT CSAIL/IBM Watson Research
Haystack:
Bringing Good Metadata to Life
Dennis Quan
[email protected]
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Outline
 Exposing the benefits of RDF data integration
 Demonstration
 Prototyping in the Haystack environment
– Hooking in different RDF sources
– Designing visualizations
– Adenine scripting language
 Example: Open Directory browser
2
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Show me the metadata
 Common questions regarding Semantic Web applications:
– “Is this stuff practical?”
– “Are you just overloading me with more information?”
– “What can I do with this data today?”
– “What is RDF giving me over databases and XML?”
 Asked by developers, not just users and observers
 Approach: easy prototyping environment for visualizing
connections within and among metadata sources
– The “museum” approach versus the “brochure” approach
3
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
The Haystack Semantic Web browser
 Allows users to create, explore, and organize RDF
information spaces
– Web browser-style navigation of Semantic Web resources
– Metadata can be fetched from a variety of sources
– User-selectable presentation templates (“views”)
– Flexible bookmark management system (“collections”)
– Access to Semantic Web Services
 Research project originating from MIT CSAIL
 Open Source Java project built on top of Eclipse, IBM’s
Open Source rich client platform
4
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Fictitious example: a rock star’s manager
 A day in the life of the manager of the famous-physicistturned-rock-star, Johnny Doe
 Some of the backend services and data have been mocked
up, but presentation services are real
– Point of demonstration is to show what can be seen through
Haystack, which is acting as a front end
 Key concepts to watch for:
– Views
– Lenses
– Collections
– Semantic Web Services
5
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
View Ontology Web Language (VOWL)
 RDF Schema, DAML+OIL, and/or OWL used to describe
ontologies to Semantic Web agents
 Similarly, VOWL is used to describe presentation knowledge
about ontologies to user agents
– Views: different ways of looking at resources
– Lenses: sets of properties that make sense being shown
together
– Operations: mini Semantic Web Services with type information
that specify what kinds of resources can be used with them
– VOWL definitions, like OWL definitions, are encoded in RDF
6
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Process diagram
Presentation
recommendations
Metadata
+
Ontological
specifications

Point and
click, hyperlinked UI
Applicable service
descriptions
7
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Incrementality in the user interface
 The more Haystack knows about an ontology, the better job it can
do presenting objects to the user
– With no knowledge, Haystack shows a property listing
– With rdfs:label and dc:title attributes, Haystack shows human-readable
names
– With rdfs:domain, rdfs:range, daml:UniqueProperty,
daml:ObjectProperty, and daml:DatatypeProperty, specialized forms
can be produced
– With lenses, Haystack shows filtered property listings in All Information
and Explore Relationships views
– With custom views defined, Haystack can show a completely custom
presentation
 These specifications do not have to all come from the same place;
different pieces of presentation knowledge can be fused together
8
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Getting metadata into the system
 Metadata can come from:
 In a number of formats:
– File system
– RDF/XML
– Web servers
– Notation3
– LSID servers
– Adenine
– Jena stores
– RSS and other XML
formats (via XSLT)
– Joseki servers
– Annotea servers
– Web Services
9
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Life Science Identifiers (LSID)
 Hyperlinking metaphor and URLs on billboards depends on
there being a metadata retrieval mechanism
 Life Sciences community coming together around LSID
– urn:lsid:[server name]:[db-specific identifier]
– Retrieval protocol based on SOAP and RDF
– Undergoing standardization by OMG and I3C
– Open Source client/server libraries provided by IBM
– Many public data sources accessible via LSID today—
beginnings of a Biological Semantic Web
 Not specific to Life Sciences
 Support built into Haystack
10
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Adenine
 Adenine is Haystack’s RDF scripting language
– Syntactically, a cross between Notation3 and Python
– Both a data definition language (RDF) and an imperative
scripting language
– Native support for RDF manipulation
– Access to Java classes and methods
 Haystack system built like a Lisp machine
– Everything is accessible from the “Adenine console”
 Leveraging the Eclipse platform
– Powerful Adenine text editor with outline and syntax
highlighting
11
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Example: an Open Directory browser
12
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Current status of prototype
 Open Source, Java/Eclipse-based implementation
 Runs on Windows, Linux, and Mac OS X
 Easy to hook in new data sources
 Stable, but still some usability issues
 Provides stable platform for extensions (Eclipse plug-ins)
Goal
Grandma
13
Haystack: Bringing Good Metadata to Life
We are here
Power user
Guru
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Packaging
 VOWL specifications can be:
– Made available for download from a Web site
– Packaged with instance metadata coming from the server
– Put into an Eclipse plug-in
 Distributing your own custom Haystack is easy
– Documentation describes process to create a strippeddown, specialized version of your own Semantic Web
browser
– Can integrate custom RDF metadata, ontologies, VOWL
specifications, and even Java and Eclipse components
14
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Real life Haystack application: myGrid provenance
Courtesy of Professor Carole Goble, University of Manchester
15
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Key ideas
 Demonstrating the value of RDF is easiest when the user can
experience the benefits for him or herself
 Haystack is an extensible Semantic Web browser:
– Connects to a variety of RDF sources
– Exposes an intuitive, Web browser-like interface
– Incrementally improves experience as more ontological and
presentation knowledge is provided
– Built on Eclipse, providing a solid basis for extensions
– Scriptable using Adenine
 Haystack addresses important HCI concerns, e.g.,
personalization and organization, that must be supported in
information applications but are often taken for granted
16
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
MIT CSAIL/IBM Watson Research
Thank you for your attention
 Dennis Quan, [email protected]
 Haystack project home page (new download coming May 24)
– http://haystack.lcs.mit.edu/
 Documentation!
– http://haystack.lcs.mit.edu/developers/
 IBM LSID home page
– http://www.ibm.com/developerworks/oss/lsid/
 Eclipse home page
– http://www.eclipse.org/
 myGrid home page
– http://www.mygrid.org.uk/
17
Haystack: Bringing Good Metadata to Life
May 22, 2004
© 2004 IBM Corporation
Descargar

How to Make a Semantic Web Browser