The annotation dilemma?

Hi,

I posted this e-mail originally on the Music Ontology mailing list. 
However, I though this topic might also be of interest on the semantic 
web mailing list.

At the beginning some background information: I co-designed the 
Association Ontology[12] over the last few weeks with the aim to be able 
to model

- association statements, which can be liked, commented etc. from other 
people
- specific categories to semantically enrich these association 
statements, e.g. mood, genre, occasion

Because, I came up with the conclusion (before starting the design of 
the Association Ontology) that there don't exist an appropriate 
ontology, which includes this already.
However, I had the feeling (afterwards) that I should give the existing 
annotation ontologies also a try to model my use cases ;)
So here we go with my results and conclusion:

========================================================================

I'd like to present association/modelling
examples with another annotation ontology. Last night, I had the time
to have a deeper look into the OAC Vocabulary from the Open Annotation
Collaboration[1] and I thought that this ontology also reflects my
aims re. annotation/association modelling. Unfortunately, the site
(the server[2]), where they described this vocabulary, was down today.
However, they wrote that their ontology is based on the Annotea
Annotation Schema with has its roots back in the year 2000. Hence, I
thought, it might be good to try this one. This schema includes a
general annotation concept for reification of the "annotates"
property, which can also be founded in several other annotation
ontologies and which is also realized by the Similarity Ontology.
Since the Annotea Annotation Schema was created in the early days of
RDF, I thought it might be good to shift this ontology and its related
Annotea Annotation Types namespace to the OWL world[5,6,7,8] (also for
testing and extension purpose).
However, when I came up to the example modelling (I took the annotated
music playlist example[9]), I noticed that there are still many
semantic relations not available. I observed that it is often the
problem in the different annotation ontologies that the developers
like to model a general applicable annotation model and thereby they
are to focused on their domain and probably lost the overview that the
annotation concept would be reutilized as extension or component in
other applications (to annotate concepts of other ontologies).
For example the Annotea Association Schema has a property called
anno:related, which should be used to related a comment, question or
whatever (the content of the annotation) to the anno:Annotation
instance (therefore I added anno:Annotation also as domain of this
property). This property was declared as to be subpropertied. That
means, every semantically richer property should a sub property of
this property and hence, also with the domain of anno:Annotation.
However, if I have a property, which is initially intended to be
directly related to something, e.g. mo:genre, and I simply want to
reuse it also in the annotation context, I can't really do this
without the application of Named Graphs, or? That was the reason, why
I kept the domain of my dcterms:subject sub properties open in the
Association Ontology.
Another example is from the OAC Vocabulary. They renamed the
properties from the Annotea Annotation Schema a bit there
(anno:hasAnnotation to oac:hasTarget and anno:body to oac:hasBody).
Furthermore, they added ranges to these annotation relation properties
(oac:Target and oac:Body). That means every thing that is used to
establish a oac:Annotation must be a oac.Target or oac:Body. Okay, I
can add this to the type description of my instances. However, this
would blew up the whole graph a bit, or? I want to reutilize existing
concepts directly for annotation statements.
Finally, I end up with copying more or less my whole association
ontology to the Annotea namespace for testing purpose and modeled then
my example[10], which now not really differs from the other example. I
more or less also end up with the conclusion that a Named Graph
(Nested Graph or whatever) based annotation/association statement
modelling approach might be the best one, because I hopefully can
reutilize, existing, semantically rich properties, without extending
their domain (to an annotation concept) and I don't have to attach
extra types to my instances to make them in an annotation/association
statement usable. In the NEPOMUK Annotation Ontology[11], they
demonstrated more or less how one can do this. However, they also used
some concepts, which are restricted to their application domain, and
they don't really aligned their ontology to DC, Review Ontology, Tag
Ontology etc.
I'm now really confused, which way I should go for solving the
annotation/association "problem".

=========================================================================

So my final question is: Am I getting something wrong or it currently a 
"problem" to model semantically rich annotation/association statements?
As far as I get through the results[13] of the Annotations Ontology 
Working Group from VoCamp 2010 they came up with more or less the same 
conclusion (use Named Graphs).

Cheers,

Bob

[1] http://www.openannotation.org/
[2] http://annotation.lanl.gov/
[3] http://www.w3.org/2000/10/annotation-ns#
[4] http://www.w3.org/2001/Annotea/
[5] http://smiy.svn.sourceforge.net/viewvc/smiy/annotea/trunk/rdf/annotea.n3
[6] 
http://smiy.svn.sourceforge.net/viewvc/smiy/annotea/trunk/rdf/annotea.owl
[7] 
http://smiy.svn.sourceforge.net/viewvc/smiy/annotea/trunk/rdf/annoteaannotationtypes.n3
[8] 
http://smiy.svn.sourceforge.net/viewvc/smiy/annotea/trunk/rdf/annoteaannotationtypes.owl
[9] http://smiy.sourceforge.net/pbo/examples/N3/playlist_-_example.n3
[10] 
http://smiy.svn.sourceforge.net/viewvc/smiy/annotea/trunk/examples/N3/anno_-_annotation_-_example.n3
[11] http://www.semanticdesktop.org/ontologies/nao/
[12] http://purl.org/ontology/ao/associationontology.html
[13] http://vocamp.org/wiki/HypiosVoCampParisMay2010#Annotations_Ontology

Received on Wednesday, 4 August 2010 10:38:36 UTC