Re: Showing the Semantic Web

Well, maybe you want a bit *more* than reified statements.  If you want 
to do statement-attribution-like things, you need more than just an 
agreed-on way to refer to the statement, you also need an agreed-on way 
to express the intended relation between the statement and some agent 
(like "said" or "asserted"), and there may be more than one.

Reification was originally intended to provide the way to refer to the 
statement, but it doesn't tell an off-the-shelf application that 
"eg:said" has attribution semantics.  As it is, Sandro's suggestion 
provides a specific way of referring to the statement, and multiple 
approaches could be developed (and might be necessary:  how many ways 
are there to refer to lines of program code, or parts of documents, or 
other parts of larger assemblages that people might want to attribute 
separately?).  But don't forget the attribution relationships themselves.

--Frank


Sandro Hawke wrote:
>>What you want is reified statements, and as Shelley Powers says,  
>>reification is "the RDF Big Ugly." There are plenty of ways to  
>>express what you want to say in RDF, but they won't necessarily be  
>>useful to a reasoner that doesn't have a built-in understanding of  
>>reification. Of course, if your target is other NewsML readers, then  
>>it may not matter, but you seem to want people to be able to use "off- 
>>the-shelf" tools to make complex queries about statements. At the  
>>moment, I don't think that's possible.
>>
>>It's too bad RDF wasn't designed with the idea that every assertion  
>>should be reified -- even if only implicitly ("foo.rdf" asserts  
>>"bar"). RDF's underlying assumption of absolute truth is a near- 
>>dealbreaker in the context of massive collaborative networks such as  
>>the internet.
> 
> 
> Or you could just use the Web.
> 
> While I'm sympathetic your complaint about RDF -- it would be nice to
> have simple reification that just worked -- we don't necessarily need it
> when we have the Web.
> 
> If you want to say 
> 
>      people:John eg:said { people:Amy foaf:knows people:Bob }.
> 
> you can instead say
> 
>      people:John eg:said <msg2231>.
> 
> and have the file (web resource) at relative URL "msg2231" contain
> 
>     people:Amy foaf:knows people:Bob.
> 
> I know it's not as elegant in some ways, but it basically works.
> 
> Now -- on the notion of absolute truth?  No, not at all.  No one says
> msg2231 is true just because it's on the web.  If the full URI is
> http://example.com/msg2231 then some people might say that Example
> Corporation is claiming msg2231, but I don't think they'd even win that
> argument.  The entities that can speak on the web -- that can be said to
> claim something -- are web pages.
> 
> I think of each URL as naming a sign-board.  The sign-board says
> something.  We can point to it and talk about.  Maybe we can tell who is
> responsible for it.  One sign board can say things about another one
> (since they all have names [URLs]).  Maybe people make directories of
> all the sign-boards (search engines).  Sometimes what's written gets
> changed.  This is the web, and if what's written is written in a
> language with declarative semantics, this is (IMHO) the semantic web.
> 
>      -- sandro
> 
> 

Received on Thursday, 23 February 2006 20:11:43 UTC