Semantic Web technologies and libraries

I read the thread about FRBR and FRSAR on the other discussion list and I'm
a little surprised by the exchanges between Karen Coyle, Jeff Young and
Herbert van de Sompel. I'm not a librarian (and more a semWeb engineer), I
work with them, that's why I may not understand all issues.
The goal of semantic Web technologies and, in particular, RDF is to offer an
interoperability framework to exchange and process data with machines in a
network world. Linked Data is the way to use semantic web technologies on
the Web with HTTP architecture. When I read the different e-mails, it seems
to me it's not clear what semantic Web technologies and, in particular, RDF
are.

Why semantic Web technologies ?
I think the key is data interoperability. Actually, the level of
interoperability for bibliographic data in libraries is the record. But, in
the web environment at Web 2.0 and mashup time, this level isn't enough any
more, we need to link data and it's very complicated to link a specific data
in the record. Imagine this use case [1] : I would like to make a mashup
between LOC data, Freebase data and RDF book mashup data. If LOC doesn't use
web technologies, I can't make it. If LOC uses a specific standard and
protocol (It's actually the case), I don't want to use them, I prefer to use
the uggly amazon data because there is an ecosystem around the Amazon Web
services. So to answer to Karen Coyle, Semantic Web technologies allow to
include library data in a networked and interoperable environment : the Web
and to take advantage of a tool/developper/standard ecosystem.

RDF is a standard model to express data thanks to web architecture (URI,
hyperlink, HTTP) and first-order logic principles. So semantic Web
technologies free the data from the record. Maybe it's just a philosophical
aspect for you, but that changes everything to make interfaces for users.
The problem isn't the use and the users, but the data, think "data first",
if you think "use first", you lock up the data. The developper/functionnal
expert/designer will have more freedom to design interfaces or find new uses
of the data if they can appropriate the data (with a common standard and a
comprehensible modelization). For example, VIAF is very interesting and the
modelling is great, but if you don't know the authority record principles,
you can't use it, even if this dataset could be useful to lots of projects
outside the libraries.

What are the RDF principles in two sentences ?
In web architecture, a URI identifies a resource. A resource is everything
we can identify : a document, a web page, a real thing... To explain RDF, I
usually use the sign theory invented by Ferdinand de Saussure : a resource
is a sign, the URI is the signifier (so an identifier AND a name) and the
real thing is the signified. RDF allows to express assertions about a sign
identified by an URI. So, a machine can use and process them to offer a
better user experience, because I can, for example, aggregate data provided
by different datasets without converting them (if they all use RDF
model...). The machine dimension is very important to understand, that's why
to have a consistent, comprehensible and usable ontology is so important
too.

>From my point of view, there are two difficulties in semantic Web
technologies :
- Think the data differently as before (triple and oriented graph model with
all consequences)
- Think the data in decentralized environment : the world wide Web. All data
are potentially connected and we have to think how describe and link data to
allow their reuse and to secure their consistency.

I don't know if this e-mail can help you, but, I needed to make it, because
in your exchange, I didn't see any more the place of semantic Web
technologies...

Best,
Gautier Poupeau

[1] In fact, I already code this mashup, see
http://www.lespetitescases.net/semweblabs/linkedbookmashup/ (in french,
sorry ;-) )

-- 
--------------------------------------------------------------------------
Gautier Poupeau
Antidot     |    GSM: +33 (0)6 45 49 59 77
F-75 Paris       |   mailto: gautier.poupeau@gmail.com
Blog : Les petites cases <http://www.lespetitescases.net> | Twitter :
@lespetitescases
--------------------------------------------------------------------------