[Top] [All Lists]

Re: [ontac-forum] What DO we want to do in ONTAC?

To: ONTAC-WG General Discussion <ontac-forum@xxxxxxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Sun, 28 May 2006 22:25:23 -0400
Message-id: <447A5B93.3020803@xxxxxxxxxxx>
Leo and Pat,    (01)

This is a very important question.  I think we agree
that it would be wonderful if we could develop suitable
resources that would support and facilitate computer
systems that could do the following:    (02)

  1. Integrate multiple knowledge sources of various kinds
     in order to facilitate the interoperability of multiple
     systems that can be, have been, or will be developed
     independently or semi-independently.    (03)

  2. Support various kinds of automated and user-assisted
     methods of search, reasoning, storage, retrieval, and
     intercommunication.    (04)

  3. Support easy-to-use interfaces that take advantage of
     the users' language abilities -- e.g., generating or
     recognizing languages that use some subset of the
     vocabulary and syntax of the users' native language.    (05)

  4. Take advantage of available resources such as terminologies,
     lexicons, thesauri, taxonomies, classifications, ontologies,
     etc., in order to achieve the goals of #1, #2, #3 above and
     any other similar goals that anyone might require.    (06)

Meanwhile, we have to recognize that very good people have been
working very hard to achieve these or related goals for the past
half century, and the landscape is littered with projects and
entire companies that have failed.    (07)

Symantec, for example, was founded in 1982 by Gary Hendrix to do
natural language processing.  Along the way, they developed some
PC utilities to support NLP. Unfortunately, the NLP goals were
never met, and the utilities became their main business.  Cyc
was founded two years later, and it survived only because of
multimillions of dollars of research funds.  They never produced
a single product or application that could pay their salaries.    (08)

Before we repeat these failures, perhaps we should ask why they
failed and whether the ONTAC goals are any more realistic.    (09)

LO> I would say ontology precedes language which precedes logic,
 > but the latter makes our understanding of the former two much
 > more precise and allows us to represent and know what we think
 > we know.    (010)

I agree that logic is useful, and I agree with Barry that Aristotle
claimed that first philosophy (which roughly corresponds to what we
now call ontology or metaphysics) somehow precedes some other kind
of philosophy.  But the people who have been doing the most work on
the subject -- that includes philosophers, linguists, and computer
scientists -- have not yet produced anything they can agree on --
not within a single discipline and certainly not among researchers
across different disciplines.    (011)

LO> ... but progress is nearly non-existent.    (012)

It's hard to make progress when there is no agreement on the
starting point, the end point, or how to get from one to the
other.  Six years elapsed since the SUO was founded.  Some SUO
groups made some progress on the goals they set for themselves,
some individuals joined one group or another, but no two SUO
groups ever collaborated to work on common goals.    (013)

PDM> ... things do not necessarily fail, they just cease to fulfill
 > a purpose and transform into something else.    (014)

That's a good point.  But then we should ask how our goals would
be transformed and what new goals would they be transformed into.    (015)

AW> ... One can then start to focus on more recent technical issues,
 > such as how to be more ambitious about robust processing of natural
 > language without falling off the "AI-complete" cliff.    (016)

It's good that you brought up the term "AI complete".  The usual
definition is that a problem X is AI complete iff X requires that
every AI problem must be solved before solving X.  Full NLP or fully
automatic computer programming seem to be AI-complete problems.
Perhaps finding a universal ontology is also one of those problems.    (017)

PC> As a working group of the Semantic Interoperability Community
 > of Practice, our focus is on finding ways to improve the ability
 > to transfer knowledge between communities that have their own
 > knowledge classifications suited to their community purposes.    (018)

But what would it mean for communities to interoperate?  A farmer
who is injured on the job might take a bus to see a physician
who prescribes some medication from the pharmacy.  But that is
very different from saying that the agricultural community is
interoperating with the transportation community, the medical
community, and the pharmaceutical community.    (019)

Interoperability among people or among computer systems is always
on a task basis.  The tasks are where they meet.  You can't develop
an ontology for enabling different systems to interoperate on some
set of tasks without examining the tasks.    (020)

PC> The list of ONTACWG objectives, from the ONTACWG home page...    (021)

But none of those objectives focus on the tasks.  They can't
succeed in making systems interoperate on some set of tasks
unless they at least look at the tasks.  For the past 15 years,
I have been telling Lenat that he should have at least some
subset of the Cyclers working on applications -- but he kept
saying that he didn't want to divert any resources from the
research.  I believe that was a major mistake.    (022)

John    (023)

Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: 
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG    (024)
<Prev in Thread] Current Thread [Next in Thread>