Chris, Bijan, Pat, All
Coming into this late, it seems to me that there are a series of use-
cases / requirements / wishes that people are trying to accommodate
made worse by real difficulties of the "use vs mention" variety. I
think there is a need of a high level document setting out the issues
separate, insofar as possible, from the nitty gritty of RDF or OWL.
Pat or others can probably help get these clearer, but s a start...
Most of this is expressed in a more OWL than RDF way because that's
how I work.
SYMBOLS, STATEMENTS & FACTS
* Reference to the specific symbols (triples) in a specific ontology
in a specific version in a specific document - required for
provenance. I have lots of requirements for tracking the changes to
reference documents at this level, where it is the symbols in the
document of a particular document that are of concern.
* Reference to the statement represented by those symbols on the
authority represented. The same statement on might have been made
elsewhere on the web with physically different symbols. Who made the
statement, etc. The symbols may change, e.g. with versioning, but
the statement persist.
* Reference to the fact asserted in the statement - e.g. whether or
not it is true, supported by sufficient evidence for a given purpose,
believed by Chris, etc.
SOME OTHER ISSUES...
GRANULARITY OF REFERENCE
* Is the class a sufficient level of granularity. In OWL I often
need to refer to specific axioms and even to conjuncts within a
definition. For example in Alan's query.
There is a subclass of gene expression processes, during each
instance of which some instance of protein a is the participant
which is "the thing produced", and which is located_in some
instance of tissue b.
Is it sufficient to refer to the entire construct? or do I need to
refer to the provenance of the claim that the protein is then located
in some issue of tissue b (rather than some tissue c)? (And is this a
"definition" in the sense of a set of jointly sufficient and
individually necessary conditions, or just a group of necessary
conditions?)
In modular OWL ontologies, the ability to annotate individual axioms
is an absolute requirement to know where they came from and should be
edited. I think I have cases for annotation and provenance where it
is needed for individual conjuncts in necessary and sufficient
conditions.
* Annotation of removal/exclusion of axioms. That an axiom is absent
is frequently as important as that it is present. When after a long
controversy an assertion is removed, I want to record that
information to help avoid somebody adding it back later.
INFERENCE AND PROVENANCE
* For OWL ontologies, we need a standard way to annotate whether a
given "statement" in OWL has been asserted or inferred. (and whether
an ontology is being provided in with all standard inferences already
performed). If it has been inferred, then we need to know the
overall scope of the inference - i.e. the set of all ontologies
imported and accessible to the reasoner during the inference. In
this case, whether or not an statement has been inferred or asserted
(by someone at some time...) is just another form of provenance
annotation.
Regards
Alan
