OWL and Published Subjects

Please find below some comments about expected interoperability of OWL,
Topic Maps in general and Published Subjects in particular.

Disclaimer : Those comments are engaging only their author, and must not be
considered as "official" feedback neither from the Topic Maps standard
working group (namely ISO/IEC JTC1/SC34-WG3), nor from the OASIS Published
Subjects TC that I happen to chair. Interoperability with OWL has not been
put on the agenda of those groups so far. That is something I personally
deplore, and that seems due to many reasons, among which: lack of task
force, "not developed here" syndrom, and last but least lack of formal
foundation for Topic Maps, making the whole TM community uneasy about the
issue.

Before going to the specific point of Published Subjects, I want to say
that I have been working for some time now, on an OWL expression for Topic
Maps, trying to consider them as some specific kind of ontology - which I
think it should be, no more, no less. Although TM gurus have always claimed
them to be ontology-neutral, this is IMO wrong, the only way to be
ontology-neutral being to keep silent. And Topic Maps folks are known to be
very talkative :)
Difficulties that I meet to achieve that task seem to come, not from the
lack of expressive power of OWL, but from the lack of consensus for what
are the foundations of Topic Maps, default a consensus on their formal
model. "What is a topic map?" is still an open issue it seems. A good
reference for the state of reflection in TM community vs RDF and OWL is the
paper from Lars Marius Garshol "Living with Topic Maps and RDF" that he
will present this week at XML Europe:
http://www.ontopia.net/topicmaps/materials/tmrdf.html.

Concerning Published Subjects, I will show below that OWL documents could
easily meet the Requirements and Recommendations for Published Subjects
expressed in Deliverable 1 of OASIS PubSubj TC [1], about to be adopted as
a TC specification (hopefully) in London next Friday:
http://www.ontopia.net/tmp/pubsubj-gentle-intro.htm.

- Requirement 1:
"A Published Subject Identifier must be a URI."

Sor far, so good. Any URI identifying an OWL class, property or individual
can be used as a Subject Identifier.

- Requirement 2:
"A Published Subject Identifier must resolve to an human-interpretable
Published Subject Indicator."

Are OWL ontologies human-interpretable? Depending on humans, of course. But
providing sensible use of labels and comments, an OWL element can be a very
accurate Subject Indicator for the class, property, or individual it
represents.

- Requirement 3:
"A Published Subject Indicator must explicitly state the unique URI that is
to be used as its Published Subject Identifier."

This is more tricky. Take for example
http://www.mondeca.com/owl/univ1_0.xml#Employee
Extracted from an attempt of re-writing a DAML university ontology in OWL)

<owl:Class rdf:ID="Employee">
  <owl:sameAs
rdf:resource="http://www.mondeca.com/owl/general1_0.xml#Employee" />
</owl:Class>

Here both http://www.mondeca.com/owl/univ1_0.xml#Employee and
http://www.mondeca.com/owl/general1_0.xml#Employee could be valid subject
identifiers. So, to be conformant to Requirement 3, either a general rule
should be set, or a specific explicit property should be set like
psi:identifier, in the following way.

<owl:Class rdf:ID="Employee">
    <owl:sameAs
rdf:resource="http://www.mondeca.com/owl/general1_0.xml#Employee" />
    <psi:identifier
rdf:resource="http://www.mondeca.com/owl/univ1_0.xml#Employee" />
</owl:Class>

This is to be discussed and settled, but the expressive power is here. Note
that the "sameAs" declaration leads to the important notion of equivalent
identifiers. A topic map application should be able to make sense of it by
aggregating on a single topic node its various subject identifiers in
various ontologies.

- Recommendation 1:
"A Published Subject Indicator should provide human-readable metadata about
itself."

This can be achieved by using specific DatatypeProperties, either at the
element level or at the ontology level, like the dc elements that can be
found in the above quoted ontology. Note that the TC has not decided yet
what kind of metadata are relevant, but there again the expressive power of
OWL can support a large variety of those.

- Recommendation 2:
"A Published Subject Indicator may provide machine-processable metadata
about itself."

No comment here. Metadata expressed in OWL are obviously
machine-processable.

- Recommendation 3:
"Metadata defined in Recommendations 1 and 2 should be consistent, but need
not necessarily be equivalent."

I figure human-readable metadata here being only a transcription of OWL
information in a more human-readable format than rdf.

- Recommendation 4:
"A Published Subject Indicator should indicate that it is intended to be
used as a PSI."

This can be specified in the ontology header.

- Recommendation 5:
A Published Subject Indicator should identify its publisher.

There again this information can be included in the ontology header.

In conclusion, I consider OWL as a very convenient format for expression of
Published Subjects. Moreover, providing the few points quoted below are
cleraly settled, it could be possible to have a recommendation for
"PSI-in-OWL", approved both by OASIS PubSubj TC as conformant to its
Requirements and Recommendations, and by OWL-WG as a recommended
application of OWL.

Again, I insist this is only so far a personal roadmap, I will put it on
the OASIS PubSubj TC agenda and see what happens there, and it's up to
OWL-WG to see if it also fits iis agenda.

Thanks for your attention

Bernard

Bernard Vatant
Senior Consultant
Knowledge Engineering
Mondeca - www.mondeca.com
bernard.vatant@mondeca.com

[1] Complete archives of PubSubj TC work are temporarily at
http://www.mondeca.com/pubsubj/
The official OASIS TC site is under reconstruction, due to CMS migration,
at:
http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=tm-pubsubj

Received on Tuesday, 6 May 2003 11:37:05 UTC