Hi Herli,
On 21/07/11 15:04, Herli Joaquim de Menezes wrote:
As matter of fact I am sure that the terminus node must be a resource.
The test ontology I made is composed by concepts linked by a simple
relationship. In the ontology these relationships are the conventional
"subclass of", and depends on, so I generated a shortest path from
ontTools in order to define a "learning path", or better a chain of
dependent concepts. So the terminus node is a concept, as the target
concept to be learned.
OK
If I understood correctly, as the "filter ANY" is used in the
pathfinding algorithm, the kind of relationship between concepts is
irrelevant, isn't it?
The path finding algorithm is a simple breadth-first search with a check
to avoid cycles. Nothing sophisticated. The filter is there to allow you
to specialise along certain types of link, so that if you were only
interested in, say, rdfs:subClassOf or skos:broader paths you could say
that.
So all things will behave as if we were walking in
a simple directed graph. I suppose this assumption is correct, although
we cannot say that a RDF graph (and OWL?) is just like a simple graph.
An RDF graph is a directed graph with cycles. The RDF representation of
an OWL model is still just RDF.
For example Hayes says the best model for RDF would be a bipartite graph
(it is in the attached file). (Is this also true for OWL graphs? I think
it is.) However as I am working with a very simple ontology, may be
these issues are not relevant.
Agreed.
So I would state the problem as:
(a) given a domain ontology how could a sequence of concepts be
extracted to be used as a learning path?
I don't know what a learning path is. Probably the best thing for you
to do is, using your test data, pick out a selection of examples of
paths that you would expect to work, and ones that you would not expect
to work. Then codify those examples as JUnit test cases, and add them to
your test suite. This has two advantages: you can be clear and precise
about what that part of your system should do, and you can ensure that
it remains correct as you develop your application.
I tried to answer this question looking for a shortest path from the
ontology as a suggestion of this "learning path".
(b) Taken as granted that this path can be extracted (I think it is
possible, I extracted one), for each concept in the path, I must find
some learning resources from a RDF metadata stored in the triple store
(as I wrote a test metadata document is in a SDB database to be used
with Joseki as a SPARQL end point).
Again, I am not sure what this means but it would be much clearer if you
could express it more specifically as one or more test cases.
Things behave as a mapping of
concepts in the path to a (finite and small) set of resources described
in a RDF matadata document. A SPARQL query is to be done on this
metadata document in SDB database to find the physical location of
resources (for instance, in a file system).
Since this system is supposed to be web based, I designed a MVC web
application so that users, for instance, in a university, could access
resources as learning material matching his/her learning strategy.
And I expect you've got some good test cases for the system overall,
don't you? Are you detecting a theme yet :)
Ian
