Adrian, (01)
When you mention Executable English, my response to you is the
same as my response to Lotfi Zadeh when he mentions fuzzy logic: (02)
It's useful for many purposes, but it's not a panacea. (03)
LZ sent a note to the BISC group (Berkeley Initiative in
Soft Computing). I decided to respond by relating his note to
the issues raised in the announcement for the CoSLI workshop. (04)
Following is my note to BISC. After that is LZ's original note. (05)
John
____________________________________________________________________ (06)
Dear Lotfi, (07)
I'd like to relate your recent note to ongoing research on language,
perception, and the computational issues of processing both: (08)
LZ> A major obstacle to precisiation of meaning is imprecision of
> natural languages. Basically, a natural language is a system for
> describing perceptions. Perceptions are intrinsically imprecise,
> reflecting the bounded ability of human sensory organs and ultimately
> the brain to resolve detail and store information. (09)
Following is an announcement of a Workshop on Computational Spatial
Language Interpretation (CoSLI): (010)
http://www.cosli.org/ (011)
An excerpt from that announcement: (012)
CoSLI> ... the same preposition can have multiple meanings, and
> such variance must be handled through either underspecified
> models that can be stretched to particular situations, or models
> which incorporate multiple disparate meanings that are assigned
> to terms as a situation invites, or models that take into account
> vague interpretations in situated contexts. (013)
This is a good summary of the issues. Before any computational
method can be applied to the problem, it's necessary to analyze
the reasons for the ambiguity and vagueness. The next sentence
of the announcement suggests some reasons: (014)
CoSLI> While early models of spatial term interpretation focused on
> the geometric interpretation of spatial language, it is now widely
> recognized that spatial term meaning is also dependent on functional
> and pragmatic features. (015)
In other words, the language a speaker uses to describe perception
is only partly determined by the geometry (What is it?) but also by
the function (What does it do?) and the pragmatics (Why do we care?). (016)
Furthermore, perception can be extremely precise. Just think of
athletes, gymnasts, and fighter pilots who accomplish amazing feats
of precise perception and response in dynamically changing situations.
Our human precision in hand-eye coordination was bred into our genes
by 30 million years of simian ancestors swinging from tree to tree.
Just one mistake could eliminate an unlucky cousin from the gene pool. (017)
LZ> What is needed for this purpose is the machinery of fuzzy logic.
> In fuzzy logic, as in natural languages, everything is or is allowed
> to be graduated, that is, be a matter of degree. (018)
As I have said many times, fuzzy logic has many useful applications.
But it does not address the critical issue of relating a time-varying
three-dimensional geometry to a linear string of words that a speaker
chooses for a particular function and purpose. (019)
The CoSLI announcement continues: (020)
CoSLI> Competent models of spatial language must thus draw on complex
> models of situated meaning, and while some early proposals exist it
> is not at all clear how geometric, functional and pragmatic features
> should be integrated in computational models of spatial language
> interpretation. (021)
I agree that your "qualitative measure of proximity" is useful for
many purposes. But much more is needed to integrate "geometric,
functional and pragmatic features" with "computational models of
spatial language interpretation." (022)
John (023)
-------- Original Message --------
Subject: [bisc-group] Representation of meaning vs. precisiation of meaning
Date: Mon, 05 Apr 2010 16:23:38 -0700
From: Lotfi A. Zadeh <zadeh@xxxxxxxxxxxxxxxxx>
To: bisc-group@xxxxxxxxxxxxxxxxxxxxxxx (024)
*********************************************************************
Berkeley Initiative in Soft Computing (BISC)
********************************************************************* (025)
Dear Members of the BISC Group: (026)
As an issue, representation of meaning has a position of centrality
in linguistics and computational linguistics. In sharp contrast,
precisiation of meaning has almost no visibility. Is there a reason? (027)
To begin with, what is meant by precisiation of meaning? The word
"precisiation" is not in a dictionary. The concept of precisiation of
meaning was introduced in my 1978 paper "PRUF--a meaning representation
language for natural languages," International Journal on Man-Machine
Studies 10, 395-460, 1978, and my 1984 paper "Precisiation of meaning
via translation into PRUF," Cognitive Constraints on Communication, L.
Vaina and J. Hintikka, (eds.), 373-402, Dordrecht: Reidel, 1984, and
developed further in subsequent papers. Today, precisiation of meaning
plays a pivotal role in Computing with Words (CW or CWW). CW is
concerned with computation and reasoning with information described in a
natural language. As we move further into the age of automation of
reasoning and decision-making, CW is certain to grow in visibility and
importance. (028)
A major obstacle to precisiation of meaning is imprecision of
natural languages. Basically, a natural language is a system for
describing perceptions. Perceptions are intrinsically imprecise,
reflecting the bounded ability of human sensory organs and ultimately
the brain to resolve detail and store information. Imprecision of
perceptions is passed on to natural languages. (029)
Theories of natural language have always been based, and continue to
be based, on bivalent logic--a logic which is intolerant of imprecision
and partiality of truth. For this reason, bivalent logic--by itself or
in combination with probability theory--is not the right logic for
dealing with imprecision of meaning. (030)
What is needed for this purpose is the machinery of fuzzy logic. In
fuzzy logic, as in natural languages, everything is or is allowed to be
graduated, that is, be a matter of degree. Bivalent-logic-based theories
are intrinsically unsuited for addressing the issue of graduation. This
explains why precisiation of meaning has almost no visibility within
linguistics and computational linguistics. (031)
Viewed as an operation, the domain of precisiation consists of
semantic entities such as propositions, predicates, questions and
commands. In the following, attention is focused on precisiation of
propositions and predicates. (032)
Let p be a semantic entity. Precisiation of p transforms p into a
proposition, p*, which is a mathematically well-defined computational
model of p. p and p* will be referred to as the precisiend and
precisiand, respectively. A basic metric of precisiation is cointension.
Cointension is a qualitative measure of the proximity of p and p*. High
cointension is associated with close proximity. Thus, precisiation is
cointensive if p and p* are in close proximity, in which case p* is a
cointensive model of p. Cointension may be viewed as a desideratum of
precisiation. Unless stated to the contrary, precisiation is assumed to
be cointensive. (033)
As a model of p, p* is assumed to be described in a modeling
(modelization) language, PL. Examples of PL are the language of
predicate logic, the language of fuzzy logic, the language of
differential equations, the language of probability theory, and their
combinations. p is precisiable with respect to PL if through the use of
PL it is possible to construct a cointensive model, p*, of p. p is
precisiable if there exists a precisiation language, PL, with respect to
which p is precisiable. Not every proposition is precisiable. (034)
Representation of meaning of p may be viewed as a step toward
precisiation of meaning of p. As a simple example, consider the
proposition p: Vera is middle-aged. The meaning of p may be represented
as: Age(Vera) is middle-age. The meaning of p is precisiated by defining
middle-age as a fuzzy set, more specifically, as a trapezoidal fuzzy
set. In this perspective, as was pointed out already, representation may
be viewed as an intermediate stage of precisiation. It is the move from
representation to precisiation that requires the use of fuzzy logic. To
substantiate this claim, I should like to pose two simple problems to
members of the BISC Group. Please use your favorite meaning
representation system, be it semantic networks, Sowa's conceptual
graphs, predicate logic, FrameNet, etc. to come up with solutions. (035)
Problem 1. Precisiate p: Carol lives in a small city near San
Francisco, with the understanding that this information would be used to
come up with an answer to the question: How far is Carol from Berkeley? (036)
Problem 2. Precisiate p: Most Swedes are tall, with the
understanding that this information will be used to answer the question:
What is the average height of Swedes? CW-based solutions will be posted
at a later point. (037)
Regards to all. (038)
Lotfi (039)
_________________________________________________________________
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 (040)
|