Gary,
Continuity in meaning is fundamental to the flexibility of natural
languages (NLs). But the formal languages of logic, mathematics,
and computer science gain precision by reducing or eliminating that
flexibility. They do so by severely restricting the range of meanings.
Since teachers use NLs to explain formal languages, it's possible
to constrain NLs to limit the meanings to a smaller set. But that
requires tight discipline and more verbose expressions.
[Kilgarriff's article is] intended for specialists in the field
of NLP [Natural Language Processing] ... The article does touch
superficially on other fields such as cognitive linguistics and
lexicography, but provides no information that would be new to
anyone who has given much attention to the philosophy of language.
The article was indeed written for readers working on or with NLP.
But Kilgarriff had a BA in philosophy, a PhD in linguistics, and
years working in linguistics and lexicography. The sentence
"I don't believe in word senses" was by Sue Atkins, who devoted
her career to lexicography, first with Collins and later at
Oxford. In 2010, Atkins and Kilgarriff founded the Master Class
in Lexicography. As for NLP, anybody working in lexicography
today uses computers to process huge volumes of data on the WWW.
you didn’t notice that [the web page] is an early chapter in
a book dealing with cenoscopic philosophy and not with NLP.
I'm aware of your project. But Atkins and Kilgarriff did not begin
their careers in NLP, and their conclusion is a corollary of Peirce's
philosophy. Like his semeiotic, it is revolutionary for lexicography,
NLP, and 20th c philosophy of language:
1. Lexicography has traditionally listed multiple word senses for
every word or term in a lexicon. The fact that dictionaries
don't agree on the number or kind of senses has been a concern
for years, but a finite set of senses for every word was always
the goal -- i.e., they hoped for polyversity.
2. For most versions of NLP, the goal is to map language to some
computable form, often some version of logic. To support that
mapping, many of them use ontologies based on the tradition from
Aristotle to Kant to the present. That tradition was strongly
influenced by philosophers, especially those with a background
in logic. They usually develop a lexicon with a finite set of
senses for each word. Polyversity would apply.
3. Anthologies with the title "Philosophy of Language" are dominated
by the mainstream of 20th c analytic philosophy. Many of them
start with something by Frege and include Russell's "On Denoting".
More recent selections are by philosophers and linguists who
use or develop representations in logic. Even if they don't use
computers, their representations can be and have been adopted by
NLP practitioners. Polyversity would apply.
The word polyversity implies that there exists a discrete set
of meanings (at most countably infinite).
GF: Nonsense. It implies variation in the relations between symbols
and their immediate objects and associated concepts; and as my chapter
says of a word, “what it denotes can vary with the user's purpose.”
As the chapter says, polyversity includes "the tendency of a meaning
to be expressible in various linguistic signs." Everyone agrees that
it's possible to have approximate synonyms, but there are few if any
exact synonyms within a single NL or between different NLs.
But we don't seem to have an established word for the tendency of
a single meaning to have many different expressions. This obviously
happens all the time; how else could you explain what a word means
by using other words?
There are many ways of teaching and explaining that don't depend
on exact synonyms: examples (special cases); examples of similar
cases with explanation of the differences; examples of opposites;
examples of generalizations... These examples could be verbal or
nonverbal (pictorial, physical, or analogous).
The possibility of exact synonyms is true of formal logics, which
have at most a countable number of denotations. Logicians often
criticize the flexibility of natural languages as "sloppy". They
want polyversity, and they deplore the lack of it in NLs.
A basic understanding of polyversity, as I call it, is much more
relevant, and does not involve any “claim that meanings are discrete”
in any formal or mathematical sense... (whatever that means).
Peirce was a mathematician. He used the term 'continuity' in its
mathematical sense: the real numbers are continuous, and the integers
are discrete. If meanings are discrete, they are analogous to the
integers. If they are continuous, they are analogous to real numbers.
This distinction is essential for understanding Peirce. If meanings
are discrete, then in any range R there is a finite number, say N.
If a word or phrase has a meaning within that range, the probability
that it is exactly synonymous with one of those meanings is 1/N.
But if meanings are continuous, then in any range of any size,
there are infinitely many. If some word W has a meaning in that
range, the probability that a paraphrase means exactly the same
as W is zero (1 divided by infinity). That implies that you might
have approximate synonyms in NLs, but not exact synonyms (except
for special cases). For examples, see slides 4 to 20 of
http://jfsowa.com/talks/natlog.pdf .
From the web page:
Maybe what i call polyversity is the same as what's called conceptual
alternativity in cognitive semantics (Talmy 2000, I.258).
No. Talmy was not talking about polysemy or its inverse. He used
the word 'window' for a kind of pattern that focuses attention on
certain aspects of a scene. Conceptual alternativity is the option
of using different patterns (windows) to focus on different, but
related aspects of a scene. Those aspects are not synonymous.
I checked the indexes of both of Talmy's volumes, and they
don't mention the words 'synonym', 'synonymous', 'polysemy'...
In fact, Talmy presented many examples of different languages
with different patterns (windows). As a result, accurate
translations between them may be difficult or impossible.
For extreme examples, see Dan Everett's work on Pirahã.
I can’t seem to avoid using that word [polyversity] ... but
Peirce didn’t avoid it either.
I believe that you misinterpreted Talmy, and I cannot believe
that Peirce meant what you're claiming.
John
