- From: Fabien Gandon <Fabien.Gandon@sophia.inria.fr>
- Date: Fri, 02 Apr 2004 16:07:39 +0200
- To: public-swbp-wg@w3.org
Hello, I spent a few days last week talking with colleagues working in the field of e-Learning. They are now moving toward Semantic Web frameworks to model and annotate their "learning objects". In the course of the discussion on metadata and curriculum modelling, several good old problems resurfaced with some additional stakes due to the SW context. I thought, may be I should reproduce some of them here as examples of genuine users' questions at very different levels of details/concerns. I might be "out of the scope" here but I believe in grounding the notes issued by this workgroup in real examples, real cases, real scenarios from real users, ... 1, "should the ontology represent the 'scientific truth' or the 'knowledge level' of the considered curriculum?" Knowing that the teachers want to share these models over the semantic web, we faced the problem of making explicit the purpose of the ontology being made available. In other words if I am sharing an ontology where whale is a subclass of fish because we modelled the current knowledge of the pupils, then an example of "good practice" (IMHO) would be to explain why and how to document these aspects; may be something in the continuation of competency questions and motivating scenarios. In other words what could be a good practice to document ontological commitments and assumptions, so that (re)users are not surprised to see Moby-Dick pop up as an answer to their query on fishes. Also what would be a good practice when linking such an ontology to other ontologies? 2, "when I want to reuse an ontology is it better to integrate it or to reference it?" What are the drawbacks and advantages of each option? Why should I take the risk of becoming dependent on an ontology I cannot "control"? 3, "how can I reconcile an ontology where possible values for an attribute are modelled as classes (e.g. <ex:very_good/>) with another ontology where the same values are modelled as individual (e.g. <ex:evaluation rdf:about="#very_good"/>)" A quick and dirty way we envisaged was to have rules translating back and forth. 4, "I found an annotation but I don't know the ontology. Should I consider the namespace URI as a URL and try to find it there? Is there an 'ontology search engine' where I can submit my URI and get the URL of a copy of the ontology?" 5, "what URI scheme should I use for instances of students, teachers, people, etc. ?" ;-) My 2 cents answer was that it does not really matter. 6, "I want to say that this person is doing 'gymnastics' and that this book is about gymnastics " i.e. this is along the lines of the debate on "classes/instances" the user needs to be able to instantiate gymnastics (a class of activities) and also to use the intension of the concept as the subject of a book (a topic). 7, "what exactly is a Resource in RDF/S?" I boldly answered "any thing that can have a URI" and ended-up (my fault) with another question "then since the output of a service has a URL (the one that triggers the call) what sense is their to annotate a changing resource (the result of the call)" And we discussed the fact that if one annotates the URL of a (frequently) changing resource then one might consider to say only things that are true whatever the content of the resource is (e.g. this resource is about weather forecast vs. this resource gives the weather forecast for the 3rd of April) or otherwise timestamp the annotation (e.g. the 1st of April, this resource was giving the weather forecast for the 3rd of April). etc. Fabien -- "Art is the lie that makes us realize truth." -- Picasso Pablo. ____________ |__ _ |_ http://www-sop.inria.fr/acacia/personnel/Fabien.Gandon/ | (_||_) INRIA Sophia Antipolis - ph# (33)(0)4 92 38 77 88
Received on Friday, 2 April 2004 09:08:40 UTC