Re: Longwell display of RDF / configuration question

From: Ryan Lee <ryanlee_at_w3.org>
Date: Thu, 12 May 2005 15:37:17 -0400

Jon Crump wrote:
> Thanks for the reply; as I read it, I realized I had stated the question
> incorrectly, though your answer was nonetheless relevant. For the
> purposes of clarification, however, let me restate:

Ah, that's what I thought you might have meant...

[snip]

> Now when longwell aggregates the two models, the Location facet shows
> that there are two records with location #china, but only one record
> with location #asia. This is because Record_A and Record_B share a
> resource <#china>, but Record_A does not have a resource <#asia>, it is
> only linked to that resource through a separate statement:
>
> <#china> <dc:title> 'China'; <dcterms:isPartOf> <#asia>.
>
> Does Record_A require an explicit <dc:location> <#asia>? Or does Record
> B need a separate statement:

Record_A needs an explicit dc:location statement.

> The problem derives from the fact that the original XML of File A
> contained separate fields for 'site' 'city' 'state/province' and
> 'nation,' thus making it possible with XSLT to generate that hierarchy
> (yes I know, I've been reading about how ill-advised I am to construct
> RDF models with XSLT. But its the tool I have, and the only one I know).
> File B, however, had no such hierarchy implied in the original XML;
> locations had to be split out of a single semi-colon delimited string.

If your data on hand is XML, then XSLT transforms to RDF certainly
aren't the worst you could do. From here, you can either rewrite your
transforms to give you those dc:location statements, or you can take
another dive into the deep end of the semantic web swimming pool and
find yourself a reasoner that would understand these two pseudo rules:

{ ?x dc:location ?y .
   ?y dcterms:isPartOf ?z . }
   implies
?x dc:location ?z .

{ ?x dcterms:isPartOf ?y .
   ?y dcterms:isPartOf ?z . }
   implies
?x dc:isPartOf ?z .

Inferencing can be a very powerful addition to your toolkit for handling
RDF. I am personally most familiar with cwm.

http://www.w3.org/2000/10/swap/doc/cwm.html

> On a separate matter. Is there documentation anywhere for exactly how
> the free-text search field works? Its behavior is a little unexpected
> and I've been trying to infer its rules of operation. It will not, for
> example, search on a string of integers, right?

I'll leave this to someone else who knows more about Lucene; the index
creation code is in model/{Local,}FacetModel.java.

What were you expecting, and what didn't meet your expectations?

> Many thanks again for your reply and to all the development folks who
> are so generous with their time.

Thanks for asking!

-- 
Ryan Lee                 ryanlee_at_w3.org
W3C Research Engineer    +1.617.253.5327
http://simile.mit.edu/
Received on Thu May 12 2005 - 19:36:09 EDT

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT