On 12 Dec 2012, at 17:44, John F Sowa wrote: (01)
> Unfortunately, I can't call in to the telecon on Thursday.
> But I'd like make some comments. I'll start with a general comment
> about Michael's list of topics: None of them mention successful
> applications of ontology to mission-critical applications.
> I am not asking for presentations *about* applications. But examples
> and use cases *derived from* practical experience are far, far more
> valuable than abstract notions of what might be theoretically useful.
> I sent a related comment to the IAOA list:
>> Suggestion: We should gather a list of *successful* cases or,
>> more specifically, *deployed* systems that implement an ontology
>> that is in daily use for mission-critical applications.
>> If we had such a list (with URLs for detailed documentation), we
>> could analyze them to determine the criteria that distinguish
>> practical applications of ontology from wishful thinking.
> This suggestion generated one response (copy below).
> I agree with Nicola on many points, but I have some serious concerns
> about some assumptions that I believe are unrealistic.
>> In the past, I have isolated and discussed such criteria: precision,
>> completeness, accuracy. See
>> Why Evaluate Ontology Technology? Because it Works!
> From Nicola's concluding slide:
>> Underspecification: simplicity encourages reusability but risks
>> to decrease interoperability.
> No! Underspecification is *essential* for interoperability. For
> examples, just look at the GoodRelations ontology, which is widely
> used and has been adopted for Schema.org by Google, Microsoft, Yahoo,
> and Yandex. That ontology defines a set of terms with a bare minimum
> of details -- and it *maximizes* interoperability. (02)
Not quite. Underspecification helps enable reuse, not necessarily
In fact, in engineering ontologies ,very specific constraints beyond what can
be stated in logic languages are often required for interoperability to be
> Precision with unique models is only possible for two kinds of
> theories: (a) very general mathematical theories, and (b) very
> specialized applications that cannot be shared with anybody else.
> Interoperability requires flexibility, *not* unique models.
> Tim Wilson asked
>> I would consider from the beginning, starting with picking a platform...
>> Can I get by with freeware like Protege or do I need the capabilities
>> of a Cyc-like application?
> Unfortunately, there is no platform for ontology that anybody with
> a mission-critical application would consider. (04)
Our TopBraid Suite platform is used every day for mission-critical apps in
numerous organisations in telecoms, Oil and Gas, pharma, life sciences, etc. (05)
> When the most widely used platform was developed by university
> students, that is a sign that the technology is very immature.
> When a platform (Cyc) can only be used by people with long training
> in AI, that is another sign that the technology is very immature. (06)
The fact that academics use Protege or Cyc says nothing at all about what
industrial organisations do or about the maturity of the technology. It's only
a comment wrt budgets vs the availability of graduate students' time and how
industrial and academic situations differ. (07)
> Matthew West has long experience in industry. He understands the
> requirements for application development, maintenance, and use:
>> We need to make sure that evaluation is grounded in meeting requirements.
>> So if we think certain evaluation criteria are important, like say
>> we need to identify a requirement such as minimising cost across multiple
> Precision with unique models would destroy flexibility, and it would
> make changes, extensions, and revisions impossible. You could not
> have multiple applications. (08)
Not quite. Take geometry in engineering design for example, where it is exactly
the details being right that enable it to be used in multiple applications
(e.g. for meshes in finite element analysis or rendering of surface finish
visualisations). It is a fair comment that less restrictive ontologies are
conceptually more reusable, however it's not always the case that restrictive
ontologies are not interoperable or reusable. The QUDT ontology is another good
example of a detailed ontology that is reusable. Luckily, languages like OWL
and SPIN also allow the layering of ontology and rule restrictions and
constraints so that they can be applied as appropriate. (09)
Based on what I've seen so far, the apparent lack of industrial experience
amongst the Ontolog community wrt what's already happening in many large
enterprises presents a problem for this particular summit. How can useful
recommendations be made by a group without that background? (010)
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2013/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2013
Community Portal: http://ontolog.cim3.net/wiki/ (012)