W3C home > Mailing lists > Public > semantic-web@w3.org > January 2007

Re: In RDF what is the best practice to represent data provenance (source)?

From: Michael Schneider <m_schnei@gmx.de>
Date: Fri, 19 Jan 2007 19:07:09 +0100
Message-ID: <45B108CD.1000002@gmx.de>
To: timbl@w3.org
CC: sandro@w3.org, richard@cyganiak.de, chris@bizer.de, semantic-web@w3.org, semantic_web@googlegroups.com

Tim Berners-Lee wrote:

> The RDF spec's version of reification is not just incomplete but broken, 
> as it does not quote the identifiers used, it uses them.  

Oh, I see: The original intention behind RDF is really to provide a 
"quoting" feature, meaning that a reified statement like

   r a rdf:Statement; subject s; predicate p; object o .

should denote the /syntactic triple/ "s p o" within an RDF graph. I now 
see it in the RDF semantics spec [1, section 3.3.1]:

   "Semantic extensions MAY limit the interpretation of these so that
   a triple of the form

     aaa rdf:type rdf:Statement .

   is true in I just when I(aaa) is a token of an RDF triple in some
   RDF document, and the three properties, when applied to such a
   denoted triple, have the same values as the respective components
   of that triple."

Apologies to Chris and Richard! I had a complete different understanding 
of reification.

Perhaps, you give me the chance to explain, how I understood reification 
until now, and then, I would like to here from you, if this alternative 
understanding is something you can better live with. Accepting this view 
on reification would, however, result in a complete re-interpretation of 
this construct.

My starting question is: How can I talk about the /relationship/ between 
two resources, which is denoted by some syntactic triple within a RDF 
graph? IMHO, the primary use of RDF in the Semantic Web is to describe 
resources, which exist in some domain. I do this by telling the
resource's properties, and by telling the relationships, which hold 
between such resources. But, I not only want to describe concrete things 
("atomic resources"), I also want to describe the relationships between 
such things. Until now, I understood reification as the means in RDF to 
do this.

 From a semantical point of view, this seemed pretty convincing to
me. RDF semantics [1] says in section 1.2:

   "The semantics treats all RDF names as expressions which denote.
   The things denoted are called 'resources', following [RFC 2396],
   but no assumptions are made here about the nature of resources;
   'resource' is treated here as synonymous with 'entity', i.e. as a
   generic term for anything in the universe of discourse."

So, for each RDF graph G, for each interpretation I of G, and for each 
URI ("name") u used within G, there must exist some resource I(u) within 
the interpreting domain, which is denoted by u. So, even for a reified 
statement r in G, like the one given above, there must exist some 
matching resource I(r) within the domain ("r" is meant to be an URI here).

The question is, what this resource actually is? What is known is that 
I(r) has the following properties:

   * the value of its property I(rdf:subject) is I(s)
   * the value of its property I(rdf:predicate) is I(p)
   * the value of its property I(rdf:object) is I(o)

The URI 'rdf:subject' would then denote that relation within the domain, 
by which we access the subject of some relationship kind of resource. 
And analog for the other two property URIs. I would further postulate 
that those three properties are restricted in a way, that they can only 
occur as properties of relationship kind of resources. Under these 
conditions, I can conclude that I(r) really must be a relationship, not 
an atomic thing. And I can also conclude that this relationship holds 
between the resources denoted by the URIs 's' and 'o', together with a 
property denoted by URI 'p'.

So, in my opinion, it is pretty natural to think of reification as a
means to represent relationships, which "live" within the interpreting 
domain, and not so much as a means for representing triples of the 
interpreted RDF graph.

> There is an outstanding RDF issue about it.
> cwm does a reification which is not broken,  using a similar but 
> different ontology.
> I think reification should be dropped from a future RDF spec.

Yes. But, with a complete change of meaning in the form proposed above, 
would you give reification a second chance? If not, can you tell me some 
alternative means to reference (and annotate) relationships within a 
domain? I do not see a second candidate for providing this functionality.

> I think n3's literal graphs like   <doc1> :says { alan :mother :joan 
> }.   should be added as an extension.
> In RDF/XML a parsetype="quote" option would do it.
> Calling them "names graphs" as caught on but they are really graph 
> literals, as they don't have to have a URI.
> To force them to have a URI would be like forcing strings, numbers, or 
> lists to haver URIs.
> Tim
> /me wonders what semantic-web@googlegroups.com is 

This is Google's Semantic Web group, a very low-frequency group, where 
this thread started originally:

   Group: http://groups.google.com/group/semantic_web

> and what its relationship is with antic-web@w3.org

I don't know about this mailing list, sorry.

Best regards,

[1] RDF Semantics: http://www.w3.org/TR/rdf-mt
Received on Friday, 19 January 2007 18:07:49 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:55 UTC