Comments on SKOS Reference

Alistair, Sean, all,

I hope everyone is having restful holidays.  

I have been studying the new draft SKOS Reference [1],
which pulls together and summarizes the design decisions
taken since Amsterdam.  Kudos to Alistair and Sean for this
substantial piece of work!

As noted in Section 1.1, some of the explanatory bits are a bit
long and should perhaps eventually be moved to the Primer, but
we need to first agree on what those explanations should say
so this is the right place to park them for now.  Telling the
story on how SKOS relates to ontologies is especially crucial.

For now, just one general comment.  

I happened to start reading the document somewhere in the
middle and found the informal prose presentation of the
model -- e.g., "skos:ConceptScheme has type Class" (4.3) --
a bit disorienting.  As a reader I found myself flipping back
to the Introduction for an explanation. Section 1.5 says that
the prose is used "to improve the overall readability of this
document, rather than mix RDF triples and other notations";
that "the meaning of this prose will be obvious to a reader
with a working knowledge of RDF and OWL"; and that "Class"
means "owl:Class", etc. This tells me enough to interpret
the above as "skos:ConceptScheme rdf:type owl:Class".  Then I
noted that there is a placeholder in Appendix C for the SKOS
data model as RDF triples.

In my recollection, we decided in Amsterdam to use prose in
order to avoid putting formulations such as

    For any resource x, all members of the set { y | <x,y>
    is in IEXT(I(skos:prefLabel)) } are RDF plain literals
    and no two members of this set share the same language tag.

into the body of the spec -- i.e., for constraints that
cannot be expressed as triples.  

I'm wondering if we really gain readability by rendering the
triples themselves as prose. Will there be many readers who
_do_ understand "has type Class" but do _not_ understand
"rdf:type owl:Class" (or simply read it as "rdf:type
rdfs:Class")?  Is it possible, on the other hand, that an
inexpert reader could get confused by formulations such as
"The Domain and Range of...", not realizing it is a shorthand
for "The Domain of..." and "The Range of..."?

My preference would be to see the components of the model,
whenever possible, in triples, and that Appendix C be
reserved for other types of formalisms (such as the above).
In order to improve readability for readers who are not used
to reading triples, the triples could perhaps be restated
as prose -- but then they could be worded in ways that are
unconstrained by the need to make the mapping obvious, e.g.,
"skos:ConceptScheme is an instance of the class owl:Class".

I suggest that everyone take the time now to give this key
new draft a careful read.

All the best in the New Year,


Thomas Baker <>
W3C Semantic Web Deployment Working Group

Received on Thursday, 3 January 2008 10:05:25 UTC