On 3/4/09 10:28 AM, "Bernhard Eversberg" <ev_at_BIBLIO.TU-BS.DE> wrote:
>> If we wish to do linking and to take advantage of data on the Web that could
>> be related to our bibliographic data, then we must change how our data is
>> stored. Names could be freely floating, if properly identified with URIs.
>
> I'd have to see working examples, end to end, to understand it. My
> imagination isn't up to this, somehow...
The idea of creating relationships between entities with URIs is very
closely tied to the concepts of the Semantic Web and the implementation is
increasingly called "linked data".
Linked data is often manifested as a stream of RDF/XML. By querying Freebase
(freebase.com) I was able to download an RDF file describing Henry David
Thoreau's book called Walden. [1] Upon closer examination of the file you
will see that it is made up of "triples" -- subjects, predicates, and
objects. For example, the subject of the whole thing is:
<fb:common.topic rdf:about="http://rdf.freebase.com/ns/en.walden">
There are then a number of predicate/object combinations. Here's one for the
author:
<fb:book.written_work.author
rdf:resource="http://rdf.freebase.com/ns/en.henry_david_thoreau"/>
Here's one for a subject:
<fb:book.written_work.subjects
rdf:resource="http://rdf.freebase.com/ns/en.house"/>
Notice how the author name and subject term are not really values but
pointers. Instead they are HTTP-actionable URIs:
* http://rdf.freebase.com/ns/en.henry_david_thoreau
* http://rdf.freebase.com/ns/en.house
Using pointers (URIs) is akin to using keys in relational databases. Instead
of inserting the value of something, you insert the pointer, and once that
is done you can change the value at the other end of the pointer instead of
all your records. No more find/replace operations. Moreover, if you are
using the Web, then the values at the other end of the pointers can be
returned depending on the language of the person do the querying or the
capabilities of the client computer.
Linked data (RDF files) have structure that is easily read by computers. It
is possible to create "graphs" illustrating the concepts represented. For
example, if you feed RDF files to the RDF Validator you can literally see
how Walden is being described. [2] If more than one item is described in an
RDF file and they share predicate/object values, then the graphs will
illustrate that. "This book is like that book in these ways."
Next-generation library catalogs should implement linked data concepts. Such
things will allow computers to create relationships between items much
faster and more accurately than humans can. These relationships might not be
"better", but they will make it easier for humans to evaluate the end
result.
Finally, the use of linked data is yet another example of how librarianship
needs to change its methods. We still need to describe materials, but we
need to do it differently.
[1] Walden - http://infomotions.com/tmp/walden.rdf
[2] Walden graph - http://tinyurl.com/8kwfad
--
Eric Lease Morgan
Head, Digital Access and Information Architecture Department
University of Notre Dame
(574) 631-8604
Received on Thu Mar 05 2009 - 21:12:28 EST