Re: NOT SPAM: a request for information!

Danny- I believe that representing facet restrictions using RDFS 
simply involves specifying the range of the properties expressing 
the facets. 

For example, consider a faceted classification of an image 
repository indexed using the Library of Congress Thesaurus of 
Graphical Materials. The TGM is comprised of two controlled 
vocabularies: TGM-I, for subject indexing, and TGM-2, for describing 
the physical characteristics of images.
 
We express the TGM-I terms as resources of type tgm1:Term and the 
TGM-II terms as resources of type tgm2:Term. Then we express a 
subject facet as a property tgm1:subject whose range is tgm1:Term 
and a material facet as a property tgm2:material whose range is 
tgm2:Term, and index the images using both facets.

To wit:

<rdf:Property rdf:about="&tgm1;subject">
  <rdfs:subPropertyOf rdf:resource="&dc;subject"/>
  <rdfs:domain rdf:resource="&imgrepos;Image"/>
  <rdfs:range rdf:resource="&tgm1;Term"/>
</rdf:Property>

<rdf:Property rdf:about="&tgm2;material">
  <rdfs:subPropertyOf rdf:resource="&dc;format"/>
  <rdfs:domain rdf:resource="&imgrepos;Image"/>
  <rdfs:range rdf:resource="&tgm2;Term"/>
</rdf:Property>

<tgm1:Term rdf:about="&tgm1;Princesses">
  <dc:title>Princesses</dc:title>
  <z3919:BT rdf:resource="&tgm1;Rulers"/>
  <z3919:BT rdf:resource="&tgm1;Women"/>
</tgm1:Term>

<tgm2:Term rdf:about="&tgm1;Oil_paintings">
  <dc:title>Oil paintings</dc:title>
  <z3919:BT rdf:resource="&tgm1;Paintings"/>
</tgm2:Term>

<imgrepos:Image rdf:about="http://www.harley.com/abstract-art/images/
(davinci)-mona-lisa.jpg">
   <dc:title>Mona Lisa</dc:title>
   <dc:description>Da Vinci's masterpiece</dc:description>
   <tgm1:subject rdf:resource="&tgm1:Princesses">
   <tgm1:subject rdf:resource="&tgm1:Forests">
   <tgm1:subject rdf:resource="&tgm1:Croplands">
   <tgm1:subject rdf:resource="&tgm1:Streams">
   <tgm2:material rdf:resource="&tgm2:Oil_paintings">
</imgrepos:Image>

Unfortunately, one has to go to Full OWL to formally express that 
tgm1:Term and tgm2:Term are disjoint classes, as in:

<rdf:Description rdf:about="&tgm1:Term">
   <owl:disjointWith rdf:resource="&tgm2:Term"/>
</rdf:Description>

But in practice, the use of URIs with lexically distinct namespaces 
enforces the disjointness. So RDFS suffices to that extent.

Of course, this begs the question of how such restrictions are 
enforced by the metadata respositories. But the representational 
machinery is largely there in RDFS to express the restriction in 
practice.

- cheers, BPA

--- In [email protected], "Danny Ayers" 
<danny666@v...> wrote:
> Hi BPA,
> 
> I'm very impressed by the work that you're doing.
> I see from the white paper that you're representing the facets as 
RDF
> properties, was wondering if/how you impose the disjointed nature 
of facets
> over and above the taxonomies. I briefly looked into expressing 
XFML in RDF,
> and found there was everything available within the RDF(S) layer 
except this
> aspect, which is available in DAML+OIL or OWL. (I was told at the 
time that
> this constraint in XFML was intended to be handled by the human 
and/or
> application rather than expressed in the language, so effectively 
XFML info
> could be expressed in RDF(S) as-is).
> 
> Cheers,
> Danny.
> 
> > -----Original Message-----
> > From: bpallen777 [mailto:bpa33@b...]
> > Sent: 13 June 2003 17:59
> > To: [email protected]
> > Subject: [facetedclassification] Re: NOT SPAM: a request for
> > information!
> >
> >
> > Bob- The amount of work depends on how much metadata is 
available to
> > the application developer, and how it lends itself to use in a
> > faceted retrieval scheme. One observation to be made based on our
> > experience is that there is far more metadata to be leveraged for
> > this kind of application than people tend to assume.
> >
> > In many commercial applications, we find that a significant 
amount
> > of effort has already been spent in record classification by
> > database designers and administrators. This existing metadata 
can be
> > leveraged in a faceted retrieval system. For example, in an
> > ecommerce catalog, products are often already classified by 
product
> > type/category, target customer, price range, etc. In our system, 
we
> > simple provide a way to automatically connect to relational 
tables
> > containing such data, transform it into an RDF representation, 
and
> > then automatically generate a navigation interface on top of the
> > unified representation. In this way, a usable faceted search 
system
> > can be generated in a few man-hours or less.
> >
> > We also find simple Dublin Core metadata, which is readily 
available
> > through many content management systems, can be fruitfully 
exploited
> > to provide faceted search over tagged content. While this is not 
a
> > faceted classification scheme in the traditional sense of a set 
of
> > facets crafted for a specific set of end-user retrieval needs, it
> > can nevertheless provide more usable navigation than is possible
> > with traditional free text or parametric retrieval approaches.
> > Coupled with named entity and noun phrase extraction techniques,
> > this can lead to a point of departure for crafting a more
> > sophisticated faceted classification scheme, while in the short 
term
> > providing real benefit to users.
> >
> > With the proliferation of XML-based approaches to content 
management
> > (particularly those associated with weblog management systems), 
the
> > amount of metadata that can be leveraged to build faceted schemes
> > will continue to expand. Again, in our system, one transforms the
> > metadata available in XML feeds and/or repositories into an RDF
> > representation and continues from there.
> >
> > In many cases, however, the intellectual effort of creating a
> > classification scheme, specifying facets and providing a 
framework
> > for a combination of human- and machine-generated metadata 
creation
> > and harvesting will need to be performed. However, our intent is 
to
> > provide tools that will, through the use of standards such as 
RDF,
> > provide a framework so that such efforts can be easily shared,
> > reused and propagated through the agency of the Semantic Web.  
Our
> > goal is to allow information architects to write RDF to create
> > working navigation systems in precisely the same way web page
> > designers write HTML to create web pages. In doing so, we can
> > make "view source" work for information architectures in a manner
> > analogous to the way it works for web pages today.
> >
> > For example, if you go to the DC-2002 example cited in my 
previous
> > post, you'll see an "RDF Metadata" button in the lower left hand
> > corner of the page. Clicking on that link will show you the 
source
> > RDF that defines the running navigation application. It contains 
a
> > simple controlled vocabulary using Z39.19 standards, a temporal
> > hierarchy, a simple classification scheme for conference events, 
and
> > the presentation metadata records. This took about a day to put
> > together based on the raw material on the conference website. By
> > providing the RDF representation directly, it now becomes 
possible
> > for others to reuse the scheme to apply it to other collections 
and
> > extend it for other uses.
> >
> > In summary, there are a number of tools and approaches that we 
use
> > to assist in the classification process:
> >
> > - Extract metadata directly from relational databases and XML 
feeds;
> > - Support incremental development of a faceted classification 
from
> > an initial set of Dublin Core metadata
> > - Use named entity and noun phrase extraction tools to create
> > subjects from unstructured content;
> > - Author, aggregate and reuse existing RDF representations of
> > vocabularies and classification schemes
> >
> > regards, BPA
> >
> >
> >
> > ----------
> > Thanks for playing. To unsubscribe from this group, send an 
email to:
> > [email protected]
> >
> >
> >
> > Your use of Yahoo! Groups is subject to 
http://docs.yahoo.com/info/terms/
> >
> >


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get A Free Psychic Reading! Your Online Answer To Life's Important Questions.
http://us.click.yahoo.com/Lj3uPC/Me7FAA/ySSFAA/0bmwlB/TM
---------------------------------------------------------------------~->

----------
Thanks for playing. To unsubscribe from this group, send an email to:
[email protected]

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/