|
|
|
Pasqualino “Titto” Assini |
|
Nesstar Ltd - UK |
|
|
|
|
The Data Web is the application of Semantic Web
techniques and principles to the problem of statistical data and metadata
(surveys, opinions polls) dissemination and processing. |
|
|
|
|
|
|
Know What You Want To Say à Modelling |
|
|
|
Then Say It -> Map the Model to a Semantic
Web Language |
|
|
|
Think of the Future à Managing Change |
|
|
|
Let Everybody Know à Publishing on the Semantic Web |
|
|
|
Make Your Entities Behave à Specifying
Operations/Services |
|
|
|
|
|
|
Keep it Simple Or Else à Why Being Too Clever
Doesn’t Work on the Internet. |
|
|
|
|
|
XML – Friend or Foe of the Semantic Web? |
|
|
|
|
|
|
|
The Semantic Web is no silver bullet. It won’t
help you in modelling your data. |
|
Use your existing modelling language (and
models). |
|
You can also use RDF/RDF Schema but: |
|
Not many (graphical tools) tools |
|
Not much expressive power (example: no
relationships) |
|
Tip: Use standard UML design tools (in most
cases a UML Class Diagram is all
you need) |
|
|
|
|
|
Once you have you model you need to translate it
to a Semantic Web Language: |
|
Choose the target language: RDF, WOL |
|
Define the mapping rules (might be already
defined, UML à RDF) |
|
What about using a specialised XML syntax? |
|
Advantage: More compact/simpler syntax |
|
Disadvantage: Difficult to design properly, need
to define an additional mapping to RDF. |
|
Tip: Stick to standard RDF syntax. |
|
|
|
|
Fact of Life: things change, models do as well. |
|
Specific Application/Clients will need to extend
your core model. |
|
Good News: RDF was custom made to be extensible
(class/property inheritance, additional properties, reification) |
|
Tip: Make sure that your applications can
handle model extensions. |
|
|
|
|
Wrong assumption: You need something special, a
Registry maybe (e.g, UDDI) |
|
|
|
This can be real easy, just do it the WWW way. |
|
|
|
Self-description principle: objects and types
self-describe themselves by making accessible, via HTTP, their RDF
description at their URL. |
|
|
|
|
Example: Bank Account operations to pay in or
draw money. |
|
|
|
The Semantic Web is silent on this point. |
|
|
|
Major oversight à Web Services come to the rescue. |
|
|
|
Problem: Web Services and the Semantic Web are
not well integrated (but we might soon have an RDF model of WSDL). |
|
|
|
Can’t wait? Use what is available: DAML-S, NEOOM (simple RDF model of methods plus
method invocation via HTML FORMs). |
|
|
|
|
Fact: The value of an information dissemination
system depends on the quantity of information it contains and the number of
eyeballs looking at it. |
|
|
|
Therefore: initially the system is worthless
(the Web was, the Semantic Web still is). |
|
|
|
But, if it starts growing it can grow real fast. |
|
|
|
So: how do you get the snowball going? KISS to
lure early adopters. |
|
|
|
|
(RDF/)XML is very hard to write by hand and not
particularly pleasant to read. |
|
|
|
It would be very useful if people (early
adopters) could jot down semantic web descriptions without tools (think of
HTML). |
|
|
|
TBL has lately engaged in an unofficial and very
politically incorrect activity: defining a non-XML syntax for RDF: N3. |
|
|
|
What about an official W3C initiative to create
a semantic web language that is not only machine understandable but also
easy for an human to read and write? |
|
Notes