[Top] [All Lists]

Re: [ontology-summit] Reasoners and the life cycle

To: Ontology Summit 2012 discussion <ontology-summit@xxxxxxxxxxxxxxxx>
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Fri, 21 Dec 2012 13:11:27 -0500
Message-id: <50D4A64F.7000006@xxxxxxxxxxx>
Alan,    (01)

I believe we agree on the issues about current systems.    (02)

> ... my concern that, if we are discussing the "ontology life cycle",
> the issues of dealing with any transformation, compilation, or inference
> step be included, whether the step uses DL reasoners or something else.    (03)

I agree.  But my claim is that good tools can revolutionize the
"ontology life cycle".  At our VivoMind company, for example, we
generate a proto-ontology directly from documents written by experts
in the field.  If standardized terminologies are available, we include
them in the mix.    (04)

But the domain experts (SMEs) don't think in terms of separate steps
of transformation, compilation, and inference.  Instead, they know
that when they flag some result as questionable, the system traces
the steps that led to the result and helps them correct the ontology
from which it was derived.    (05)

I admit that these tools are still in the experimental stage, but
we have used them for deployed applications.    (06)

I'll quote the concluding sentence of the following note:    (07)

> It's a lot cheaper to put some R & D money into developing better
> tools than to train domain experts to become knowledge engineers.    (08)

And by the way, I don't want to look like I'm just badmouthing OWL,
but the OWL restriction to tree-structured models is not helpful.
It makes the logic much more quirky with too many restrictions.
That is an obstacle to developing automated tools.    (09)

John    (010)

-------- Original Message --------
Subject: Re: [ontolog-forum] Ontologies and individuals
Date: Fri, 21 Dec 2012 12:22:30 -0500
From: John F Sowa    (011)

Dear Matthew, Pat, Paul, and Doug,    (012)

Fundamental principle:  Our systems must be able to support diversity,
heterogeneity, and interoperability.  That means they must be very
general and flexible.  But we also need good tools to support people
at every level from abstruse theoreticians to typical iPhone users.    (013)

> That is a "do whatever you like" approach. It's fine for one person, but
> it just leads to chaos if you have say hundreds of people working on the
> development of an ontology.    (014)

Pat was trying to say that the underlying formalism must be sufficiently
flexible to support anything that any system might need.  But that does
not rule out the option of restricting the flexibility in order to
standardize the choices for a particular system or application.    (015)

> ... should we banish all terms that engender argument (as John Sowa
> suggests), or those that fill out the upper levels of a taxonomy or
> require interpretation by a guru (as you [Pat] suggest)    (016)

You don't have to banish a term just because somebody raises
a quibble about it.  But when you hit a deadlock about some term
with dozens of notes by knowledgeable people and no consensus
among them, you might realize that the term is too vague, ambiguous,
or problematical to be useful.    (017)

Re abstruse terms:  Many aspects of system design require people
with a high level of expertise.  But the user interfaces must
use appropriate terminology designed for the different kinds of
people who use it.    (018)

In particular, experts in medicine or geology would use common
terms in their field.  They would never use technical terms
in logic or ontology.    (019)

> But the human users and developers do not just give an
> ontology to a logic engine and let it have its merry way.    (020)

Of course not.  You have to design appropriate tools.    (021)

As an example, every implementation of SQL since the 1970s
implements a "knowledge compiler" of sorts.  The WHERE clause
in SQL uses full FOL, but a simple-minded translation of that
clause to an executable form would be horribly inefficient.    (022)

Prolog forces the programmer to do the optimization by
reordering the rules.  But you can't expect SQL users to
have the same level of expertise as Prolog users.  Therefore,
SQL implementations do the reordering when they translate SQL
to an executable form.    (023)

> The programmers use the comment fields (and often term names)
> in order to make such connections between the ontologies and
> the real world.    (024)

Yes, but the computer cannot understand the comment fields.    (025)

There are tools, some in operational use, that can do much more.
It's a lot cheaper to put some R & D money into developing better
tools than to train domain experts to become knowledge engineers.    (026)

John    (027)

Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/   
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/  
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
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/     (028)
<Prev in Thread] Current Thread [Next in Thread>