Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: SUO: Automated or Semiautomated Ontology Development




On 3/10/02 11:30, "John F. Sowa" <sowa@bestweb.net> wrote:

> Today, I happened to read two passages that reinforce one another in a way
> that suggests a better way of doing ontology as well as many other tasks in
> knowledge representation.  Neither of the topics is new, and I'm sure that
> many, if not most readers of these email lists have come across them in one
> form or another.  But together, they emphasize an important point.

John, Jon, et al..  A number of things to say here:

1. On the "What is" issue

  a) It assumes that some preoccupation with "what is" is present in this
  group.  Some in this group are, to a greater or lesser extent, concerned
  with what is.  This involves the belief that there is some objective
  reality to be gotten at.  I personally believe this is likely to be the
  case.  It also involves the epistemological issue of whether if there is
  a "what is" that we can get at it.  Well, some think we can, and some
  think we can't.  So what?

  b) Presence of bickering is no evidence against the thesis.  In fact we
  could say the same thing about the program being advocated by JA & JS.
  Once we stop caring about "what is" and start caring about what people
  say about "what is" we're in no better shape than before.  In fact we're
  in worse shape.  Why?  Imagine putting the same requirement on the
  physical sciences.  Perhaps we should stop worrying about all those pesky
  particles and fields and just worry about aether and phlogiston.  After
  all, those were ways entirely reasonable people talked about "what is",
  right?

  c) If you don't want to worry about "what is", then just adopt the
  "Gruber" definition of "ontology" where one simply worries about the
  expression of "conceptualizations" (I'd guess you'd accept those as being
  ways to talk about "what is"") in formal languages.  But then you have a
  burden to bear - a lot of the return to metaphysical talk in the ontology
  community, as championed by Nicola Guarino, Barry Smith, Aldo Gangemi,
  and others, is a response to the manifest inadequacy of the attempt to
  simply axiomatize commonsense as an answer to the problem of providing a
  universal substrate for information exchange on the web and elsewhere.

2. On Google

  Google's ontology consists of two categories: node and link, hooked up to
  a boolean keyword search.  For some searches it does remarkably well; for
  others, not.  IMHO, they've pushed the IR paradigm about as far as it can
  be pushed.  It's brilliant, but anything that goes further is going to
  have to start looking into NLP. Is this situation going to be any better
  off than the search for the metaphysical holy grail?  I don't think so.

3. On automated "ontology" construction

  We already have this - it's called dictionary construction and it's what
  lexicographers do for a living.  In fact, WORDNET should be a paradigm
  case for what you want.  Now, nice at it is, it is not being heralded as
  the solution to everyone's information exchange problem.

  Again, if you (meaning anyone out there who advocates the same course as
  JA and JS seem to be taking here) think that this will work, then I will
  say the same thing that was said to Chris Lofting about his ideas a while
  back: 

    Step up to the plate and write the code, and we'll see if it works.

  .bill