[WN] Draft review

Hello,

1. The problem statement implicit in the Introduction is not clear enough. The
Introduction states that WordNet "has also been adopted in the Semantic Web
research community for use in annotation, reasoning, and as background knowledge
in ontology mapping tools" (please provide references for all this cited work,
and are they the same 'developers' mentioned by "The WordNet Task Force of the
SWBPD WG aims at providing a standard conversion of WordNet as a reference point
for developers").

Is the problem that these efforts are not well coordinated BECAUSE each is
working with  unmappable and incompatible, representations of the WN database?
Why is a single model important? It's unclear who the intended readers are for
this Note -- is the 'Semantic Web research community' the (only) stakeholder?

2. The Introduction doesn't identify what 'end-product' is envisioned by this
Note -- is it a statement of Best Practices, or is it actually a conforming
database? If the latter, the note likely should address operational policies
like versioning and access. Personally I think it's best to leave conversion to
others, the purpose of this note being a consensus statement by stakeholders.

3. The "WordNet Datamodel" section deserves more substance than it now has. This
section is important because it presents the consensus about what data is
represented within the WN database. Succinct definitions of each WN term to be
modeled in an RDFS/OWL ontology, are most definitely needed here.

4. The "Conversion to RDF/OWL" section could be transitioned to a more
comprehensive analysis of each of the three conversions accomplished to-date.
Their data models could be better described, so as to identify what information
now supplied by one or another will (not) be accommodated by the Note's own WN
ontology.

5. A "Design Approach" section needs to forthrightly address whether the WN
database contains conceptual entries or ontology entries. My understanding is
that if a taxonomy is not an ontology, then it is a concept hierarchy. In other
words, please answer this: why is WN (not) a natural SKOS concept hierarchy? Of
course, a SKOS concept hierarchy would be annotated with WN-specific information
e.g., pointers to antonyms. In other words, is synset a subclassOf a
skos:concept, and why (not)?

6. A "WordNet Ontology" section could document -- for each and every term in
"WordNet Datamodel" -- pertinent RDFS and OWL definitions. It could address the
impact of the three OWL processing models on the applicability of each
definition.

7. A "WordNet Database Conformance" section could discuss minimum requirements
that a WN database purveyor would need to meet. Here one would expect
non-trivial sample entries to be provided.

8. "Using WordNet Databases" could non-normatively discuss what value-adds might
ultimately be developed by stakeholders including: a) enhancements to the
database and b) tools that use or rely upon accessing a conforming database.

Should SKOS be the basis for representation of the WN database (and I think it
should be) then it'd be quite useful to discuss whether and how a definition of
a WN entry in some ontology can be related to its definition as a SKOS concept,
that is, to allow one to discover ontologies that define a WN concept as an
rdfs:Class or owl:Class rather than as a skos:concept.

Regards,
John McClure

Received on Monday, 28 November 2005 21:43:02 UTC