[Top] [All Lists]

Re: [ontolog-forum] Terminology and Knowledge Engineering

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ed Barkmeyer <edbark@xxxxxxxx>
Date: Tue, 24 Jan 2012 18:50:06 -0500
Message-id: <4F1F43AE.9030503@xxxxxxxx>

John F. Sowa wrote:
> I came across the ISO standard 704
> for "Terminology work - Principles and methods".  ...
> That document has some interesting examples, but there are no hints
> of any formal notations.  It shows why the terminology of a field
> is important as a prerequisite for an ontology of that field.  It
> also shows why there is a large gap between terminology and what
> the theoreticians say about formal ontology.
>       (01)

Remember first that the audience for that work is people who write 
domain vocabularies, standards of practice and other semi-formal 
publications.  The objective is to explain to such persons how to select 
terms, and conceive and write definitions that will be clear and 
understandable to other people who have some training in the subject domain.    (02)

Contrast that with the idea of formal ontology, which is about the exact 
specification and use of terms in documents intended to be consumed by 
automata.  The point of the ontologies is to ensure some mathematical 
rigor in the specification, because mathematical methods will be 
involved in its interpretation and use.  That is a very different objective.    (03)

That said, as of the turn of the century, there has been considerable 
motion in the 'terminology' communities toward formalizing terminologies 
as ontologies, as the automata become more intimate partners with humans 
in finding information and making decisions.  It is important to get the 
people and the machines "on the same page".  ISO 704 is 30 years old and 
in its 5th or 6th revision.  The world is moving on, and the TKE 
conference is a demonstration that even the old guard of ISO TC37 (the 
body that produced ISO 704) are aware of the need to move forward in 
knowledge engineering for their trade.    (04)

> But I also noticed that the gap between terminology and what a very
> large number of OWL practitioners do is much, much smaller.  In fact,
> the overwhelming number of OWL "ontologies" published on the WWW just
> take the informal info from a terminology, put angle brackets around
> it, and call it an ontology.
>       (05)

There is certainly truth to that, but as others have observed, just the 
limited formalism of defining classes and properties (nouns and verbs) 
in a formal way is a step forward.  It standardizes machine readability 
of the vocabulary terms, and it associates them with some resource that 
documents their intent, for the benefit of the human users who will be 
working with the machines that to the reading.    (06)

Further, this is 'crawl before you walk'.  Just getting this degree of 
formalism is breaking a barrier.  It is making the domain experts aware 
that there is information technology that supports what they do, and 
they don't need an advanced degree in mathematics or computer science 
(as distinct from government, or finance, or automotive engineering, or 
habitat management, or medicine) to use it.  It should serve to 
demonstrate that our fine AI technologies can actually be useful to 
them.  Their successors will realize that they can add property 
definitions and axioms to the ontology, and those additions will improve 
its usefulness.    (07)

Finally, professional knowledge engineers tend to push OWL to its limits 
in supporting automated reasoning.  OWL effectively embodies a number of 
axiom schemata that support common expressiveness needs.  The problem is 
that the resulting OWL models are no longer human-readable, or at least 
not in the area of the add-on axioms.  Most people readily grasp simple 
subsumption, but property-defined classes can be (in my experience) 
rather difficult for domain experts to understand conceptually, to say 
nothing of the ugliness of the OWL representation.  (I remember having 
to explain the distinction between necessary and sufficient properties 
twice, once to a business rules body and once to an engineering 
standards body.  And I have also presented formalized English sentences 
together with the CLIF form generated by a compiler.  The reaction of 
the audience was: I understand the 'English' and I'm sure glad I don't 
have to try to understand CLIF.)  What this means is that the more 
formal and rich the ontology is, the less accessible it is to the people 
whose knowledge was used to create it, and who will be the partners of 
the automata that use it.    (08)

So, I personally think that the crude ontologies are a demonstration of 
'virtus in medio'.  There is a lot of current value to the baby steps in 
generating domain ontologies.  There is a whole community that is coming 
to understand what we do at a basic level, and to see what we do as 
valuable.  We need to encourage them, not disparage their work, or 
present them with advanced knowledge engineering stuff that will drive 
them away screaming.  It won't do to be the 'popinjay bravely born, who 
turned up his noble nose with scorn at the loving heart that he did not 
prize...'*    (09)

> Since I only attended one terminology conference many years ago, I
> would like to ask any Ontolog subscribers who may have more experience
> with the field about their views of the relationship between terminology
> and ontology.
>       (010)

We are starting to see several standards for managing 'vocabularies' as 
information repositories, e.g., SKOS, SBVR, the draft ISO 11179-3, the 
draft ISO 29002, etc.  Of these, only SKOS and SBVR understand the idea 
of 'definition in a formal language'.  For SKOS, of course, the language 
is RDF, and the SBVR lot invented their own.  The others all have the 
idea of definition in multiple languages, but they don't distinguish 
natural language text from formal language syntax, because much of their 
community concern is having definitions in (e.g.) English and Chinese (a 
more traditional 'vocabulary' concern).     (011)

It seems to me that an ISO 11179 'terminology' could easily be converted 
to a crude OWL model, with the URI referring to the ISO 11179 
repository.  So, the distance is exactly what John described above.  By 
comparison, a SKOS vocabulary _is_ an ontology -- it has a standard RDF 
form.  Whether it is 'crude' or not depends on whether RDF definitions 
are common in it.  In a similar way, an SBVR 'vocabulary' (or one of its 
several packaging concepts) could be converted to an ontology, and the 
result will be a crude ontology if all the terms have only English 
definitions, or a genuinely rich ontology if most of the definitions 
have what they call 'semantic formulations' in their XML representation 
of logic.  (Their logic includes modalities and nominalization and 
Davidsonian treatment of situations, but it has no formal semantics -- 
it is a grab bag of logic extensions.  OTOH, a reasonable subset has a 
mapping to CLIF.)  And, unsurprisingly, many of the founding SKOS folk 
and SBVR folk and a few ISO 11179 folk are co-workers with the 
vocabulary standards folks in ISO TC37.  So, it is my assessment -- 
admittedly from the particularly warped lens of standards work -- that 
some part of the 'vocabulary' expertise and its products already have 
one foot in ontologies.  And that is a good thing.    (012)

-Ed    (013)

* from "I have a song to sing, O!", W.S. Gilbert
> I would also like to ask OWL advocates about those popular ontologies
> whose only "definitions" are English phrases marked as comments.  What,
> if anything, do they get from such an ontology that goes beyond what
> they could get from a well-written terminology?
> John
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
> Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/ 
> To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
>       (014)

Edward J. Barkmeyer                        Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263                Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263                Cel: +1 240-672-5800    (015)

"The opinions expressed above do not reflect consensus of NIST, 
 and have not been reviewed by any Government authority."    (016)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J    (017)

<Prev in Thread] Current Thread [Next in Thread>