[Top] [All Lists]

Re: [ontolog-forum] Danger of URIs in mission-critical applications

To: "'[ontolog-forum] '" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "Patrick Cassidy" <pat@xxxxxxxxx>
Date: Fri, 10 Jul 2009 18:11:54 -0400
Message-id: <0ed201ca01ab$6eeafdd0$4cc0f970$@com>
Rich Cooper asks:    (01)

> Pat, with these clearly stated goals, I can see why the FO would
> require a
> common vocabulary, syntax, even simple semantics for representing the
> lattice of theories, even when we don't know what the lattice structure
> is.
> But why a full ontology?  Why is the rest necessary?
> And where can I download one?    (02)

   The common foundation ontology at a minimum should represent all of the
semantic primitives required to describe anything any of its users want to
represent.  It is likely to increase in size until some number of different
domain ontologies have been defined in its terms. The point is that a common
FO is most effective if anything one wants to represent can be represented
as an FOL combination of the terms in the FO; then, semantic
interoperability only requires that one communicate, along with data, the
FOL definitions of terms not already in the FO - those can be properly
interpreted as long as the FO terms are properly interpreted.  The
'primitive' representations in the FO will be those that cannot be
represented as FOL combinations of other FOL terms - for example, functions
that require procedural code (arithmetic functions, graphic output).  Other
'primitives' would be terms that can only be understood properly by knowing
instances.  I suspect that human feelings and emotions would also be among
those that cannot be adequately represented by terms in any language.  From
various means of estimation, I think that such an ontology would have at
least 6000 and possibly up to 10,000 elements (types, relations, functions,
rules) . .  and maybe more - the number of primitives needed is a question
that can only be answered by experiment.
   But in addition to the true primitives, it would be very helpful to also
include representations of concepts that are uncontroversial and
sufficiently common that they make a natural language interface more
effective.  Based on our experience with the rate of adoption of existing
upper ontologies, it seems likely that no foundation ontology will become
widely used unless it has some effective natural language interface, that
can take a description of a concept in, e.g. English, and answer whether
that concept or something very close is already in the ontology, and if not,
to create the proper representation of the concept described.  This is
likely because, with its ability to represent any concept at all, the FO
will have the complexity of a basic human language (sign language
dictionaries have from 2,000 to 5,000 signs).  There may well be specialized
extensions with fewer concepts that would be easier to learn how to use
effectively.  I also have in mind a goal to develop a NL program with the
ability to converse naturally with a 6-year-old.  That could be an extension
of the NL interface that would also serve as an example of how to use the
ontology in applications.    (03)

[RC] > And where can I download one?
[[PC]] There are several 'upper ontologies' that could serve as a starting
point for a common FO, but none have yet been adopted widely enough to serve
the purpose. In my own inspection, none of the existing 'upper ontologies'
have representations of all of the words in the Longman dictionary defining
vocabulary, which I am taking as a first estimate of the inventory of
primitives required to perform the function of a 'conceptual defining
vocabulary' when represented as an ontology (some of the words have several
basic senses).  I am in the process of creating a foundation ontology that
does include representations of all of the Longman words, using the existing
FO's (mostly OpenCyc) to supply worked-out ontology elements, and
supplementing where needed.  The COSMO ontology is being first developed in
OWL, but needs to have an FOL (CL-compliant) representation before it will
be adequate to the task.  COSMO is available at http://micra.com/COSMO.    (04)

   In order to function to support interoperability the FO has to be a
"common FO", both fully open-source, and used and maintained by more than a
few groups.  It seems likely to me that a widely used FO will only develop
as the product of an organized development effort that involves over a
hundred participants, to ensure that all of the primitive elements required
to define concepts in any domain will be identified.  The FO itself will
ideally be stable so as not to risk changes that cause errors in existing
applications; but there can be any number of domain-specific extensions,
including those that have mutually inconsistent assumptions.  The biggest
problem is getting enough funding to execute such a collaborative
development (ca. $30M).  It may take a while.  Meanwhile, I am developing
the COSMO as a first approximation to an adequate FO, to allow testing the
question of just how many 'primitive' concepts are needed to represent
things in a large variety of domains.  It is still in progress and
incomplete - there are over 400 Longman words yet to be represented, and it
is still only in OWL, not CL.    (05)

Pat    (06)

Patrick Cassidy
cell: 908-565-4053
cassidy@xxxxxxxxx    (07)

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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (08)

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