Re: "Hard" categorisation cannot do
Jean-Luc,
Thanks for the reference to Louise Connell's thesis:
http://wombats.ucd.ie/papers/Thesis2000.pdf
She makes some interesting observations, and I
certainly agree with her point that categorization
is highly context dependent. I've been saying that
for years in my papers on knowledge soup and related
issues.
I also agree, to a certain extent, even with her
opening quotation by Bertrand Russell:
“What, exactly is meant by the word ‘category’,
whether in Aristotle or in Kant and Hegel, I must
confess that I have never been able to understand.”
– Bertrand Russell: History of Western Philosophy
As I said before, Russell's book is the last place to
look for a sympathetic reading of anybody's philosophy
other than his own. But I agree with him the word
"category" has been bandied about in a very loose way
by many people.
However, Aristotle and the medieval Scholastics used
it in a very precise way, which is indicated in the
etymology: Aristotle used his word "kategoria" for
the general terms in his syllogisms. The Scholastics
translated that word to the Latin "praedicatum",
from which we get the predicates of predicate calculus.
I use the term as a synonym for a monadic predicate,
especially when it is used to classify entities.
That is the classical sense of the term, and it is
the only one that anybody has been able to define
precisely.
Given that definition, it is certainly true, as Louise C.
observes, that no predifined category (or predicate) is
likely to correspond with the context-dependent clusters
that are derived from a statistical analysis of any text.
But there is a lot more to say about the subject, and I
would recommend a book by Alan Cruse, _Meaning in Language:
An Introduction to Semantics and Pragmatics_, Second Edition,
Oxford University Press, 2004. Following is a web site
that summarizes some points from the book:
http://home.tnv.mh.se/mats-deutschmann/Summary%20of%20Cruse.pdf
In particular, I like Cruse's term "microsense" for the sense
of a word in any particular context. His point, which I
wholeheartedly endorse, is that the small number of word senses
found in typical dictionaries are merely rough groupings of
the open-ended number of microsenses that a word can have in
each context of its use.
To relate that to the term category: any word in any natural
language has an open-ended number of microsenses, each of
which may be defined by a different category or predicate.
JLD> Nitpicking about CG versus DL versus whatever might be
> pretty irrelevant...
Of course. It has no relevance whatever to this topic.
And by the way, there is no "versus" between CGs and DL.
As Enrico and I agreed, you can use DL techniques very
nicely within a CG system.
John