W3C home > Mailing lists > Public > semantic-web@w3.org > September 2010

Time for quintuples?

From: Bob Ferris <zazi@elbklang.net>
Date: Thu, 23 Sep 2010 20:43:24 +0200
Message-ID: <4C9B9FCC.5000008@elbklang.net>
To: semantic-web@w3.org
Hello everybody,

I thought recently (again) about the handling of detailed descriptions 
of an triple and hence reification. I followed in the last time often 
the approach of object-oriented context reification. That means, I 
introduced a new concept to describe a relationship more in detail. This 
approach works fine if there exist no property hierarchy.
However, I designed exactly an ontology[1] with this feature. The used 
property there is cco:cognitive_characteristic[2], which has further 
specific sub properties, and the applied reification concept is 
cco:CognitiveCharacteristic[3].
Initially, this design should enable to define different cognitive 
patterns of an agent, which have the same topic as object, e.g. one is 
_interested_ in soccer, has some _skills_ in soccer and also some 
_expertise_ in soccer. With the Cognitive Characteristics Ontology it is 
possible to simple represent these statements as triples, e.g.

ex:APerson
	a foaf:Person ;
	foaf:name "John Wayne" ;
	cco:skill <http://dbpedia.org/resource/Football_(soccer)> ;
	cco:expertise <http://dbpedia.org/resource/Football_(soccer)> ;
	cco:interest <http://dbpedia.org/resource/Football_(soccer)> .

However, these simple statements do not say anything about the levels or 
weightings of these cognitive patterns of this person, rather then 
something about the related activity, e.g. playing or watching, or 
characteristic dynamics. Hence, one can use the 
cco:CognitiveCharacteristic concept to represent this knowledge, e.g.

	cco:habit [
		a cco:CognitiveCharacteristic ;
		cco:topic <http://dbpedia.org/resource/Football_(soccer)> ;
		wo:weight [
			a wo:Weight ;
       			wo:weight_value 6.0 ;
       			wo:scale ex:AScale
       			] ;
       		cco:activity 
<http://sw.opencyc.org/concept/Mx4rwJRiEpwpEbGdrcN5Y29ycA>
		] ;
	cco:habit [
		a cco:CognitiveCharacteristic ;
		cco:topic <http://dbpedia.org/resource/Football_(soccer)> ;
		wo:weight [
			a wo:Weight ;
       			wo:weight_value 7.0 ;
       			wo:scale ex:AScale
       			] ;
		] ;
	cco:habit [
		a cco:CognitiveCharacteristic ;
		cco:topic <http://dbpedia.org/resource/Football_(soccer)> ;
		wo:weight [
			a wo:Weight ;
       			wo:weight_value 5.0 ;
       			wo:scale ex:AScale
       		] ;
       		cco:activity 
<http://sw.opencyc.org/concept/Mx4rwO0J55wpEbGdrcN5Y29ycA> .
		] .

ex:AScale a wo:Scale ;
	wo:min_weight 0.0 ;
	wo:max_weight 9.0 ;
	wo:step_size 1.0 .

Unfortunately, these detailed descriptions aren't related to the 
intended cognitive pattern. In a simple use case on can match this 
relation via the cco:topic property, which should have the same topic as 
the cognitive pattern. However, this wouldn't work here.
Hence, we will need a mechanism, which binds the reification statement 
to its triple. I tried here three different variants:

1. Named Graphs, where every triple and its related reification 
statement is entailed in a separate graph (see [4])

2. Named Graphs, where every triple that should have a reification 
statement is entailed in a separate graph, and the reification statement 
is also the graph description, hence, type of both - rdfg:Graph and 
cco:CognitiveCharacteristic (see [5])

3. N-Quads, where the reification statement is referred via the context 
node (see [6])

This result let me also think about the quintuple approach again. That 
means, to be more concrete: a combination of the Named Graph and the 
N-Quad approach, where the context node of the N-Quad represents the 
reification statement of the relation represented by the related triple 
and the surrounding Named Graph represents the "common" provenance and 
trust information (as these use cases are often proposed as common for 
Named Graphs). Would you agree with that modelling?
For example:

ex:NG1 { ex:APerson
	cco:skill <http://dbpedia.org/resource/Football_(soccer)> ex:CC1 . }

ex:CC1
	a cco:CognitiveCharacteristic ;
	... .

ex:NG1
	a rdfg:Graph ;
	dcterms:modified "2010-09-22T09:55:52+01:00"^^xsd:dateTime .

This would also make the explicitly modelling of the reification triples 
(via rdfs:subject, rdfs:predicate and rdfs:object) in the reification 
statement, as proposed in the RDF Reification of the RDF Semantics[7], 
obsolete.

Cheers,


Bob


[1] http://purl.org/ontology/cco/cognitivecharacteristics.html
[2] 
http://purl.org/ontology/cco/cognitivecharacteristics.html#cognitive_characteristic
[3] 
http://purl.org/ontology/cco/cognitivecharacteristics.html#CognitiveCharacteristic
[4] http://smiy.sourceforge.net/cco/examples/N3/cco_-_football_example.trig
[5] 
http://smiy.sourceforge.net/cco/examples/N3/cco_-_football_example_02.trig
[6] http://smiy.sourceforge.net/cco/examples/N3/cco_-_football_example.nq
[7] http://www.w3.org/TR/rdf-mt/#Reif
Received on Thursday, 23 September 2010 18:43:56 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:38 GMT