W3C home > Mailing lists > Public > www-rdf-interest@w3.org > February 2005

Re: missing bit of RDF for XML people

From: Frank Manola <fmanola@acm.org>
Date: Thu, 03 Feb 2005 10:42:40 -0500
Message-ID: <42024670.1010901@acm.org>
To: Stefano Mazzocchi <stefanom@mit.edu>
CC: www-rdf-interest@w3.org

I (obviously) don't disagree with your general objective, but it seemed 
to me the original question had to do with capturing the XML that was 
there in as direct a manner as possible;  as Karl said, "without 
thinking about writing good RDF".  Danny's reference to the infoset was 
also relevant to this point.  I also don't think something like this is 
necessarily "five steps backwards", nor is it "completely useless". 
Clearly, further processing is needed to get "good" RDF, but you're no 
worse off with it than you were before are you?  And you can do your 
further processing entirely on RDF, considering the stuff below as an 
intermediate representation.  Whether doing it that way is preferable to 
doing more direct transformations on the XML is, of course, another 
issue.  It depends to some extent on what kinds of descriptions of the 
appropriate transformatons you have available.

--Frank

Stefano Mazzocchi wrote:
> Frank Manola wrote:
> 
>>
>> You don't necessarily need a *blank* predicate.  You just need a 
>> "standard" predicate (standard for somebody!) that means "is related 
>> to" (in some way that you haven't identified further).  Using this, 
>> you'd write something like:
>>
>>   subject        predicate          object
>>
>>  country  - ex:isRelatedTo --> "Canada"
>>  country  - ex:isRelatedTo --> city  - ex:isRelatedTo --> "Montréal"
>>  country  - ex:isRelatedTo --> city  - ex:isRelatedTo --> "Toronto"
>>  country  - ex:isRelatedTo --> city  - ex:isRelatedTo -->
>>  "Vancouver"
> 
> 
> I strongly believe this would be one step forward and five steps 
> backwards. The above might be valid RDF but it's completely useless: 
> throw it into a triple-store and it feels just like a lot of noisy bNodes.
> 
> I would much rather have some sort of declarative 'semanticsheet' 
> (produced by the data makers or by some 'semantic annotators') that 
> makes the information implicitly encoded in the xpaths explicit:
> 
> country/text() -> dc:title
> country/city -> geo:city
> 
> but this is only half of the puzzle, since the URI identification of the 
> nodes is still missing.
> 
> [BTW, this is part of my research work on the RDFification of data for 
> the SIMILE project, the above 'semanticsheet' could be easily translated 
> into XSLT]
> 
Received on Thursday, 3 February 2005 15:35:31 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:13 GMT