Professional Documents
Culture Documents
zadeh_preceison in computation
zadeh_preceison in computation
zadeh_preceison in computation
Lotfi A. Zadeh
Computer Science Division
Department of EECS
UC Berkeley
z Precisiation of meaning is a
prerequisite to computation with
information described in a natural
language. In turn, understanding of
meaning is a prerequisite to
precisiation of meaning.
5 /168
IMPRECISION OF NATURAL LANGUAGES
description evocation
perception NL perceptions
z p: perception
z NL(p): description of p; semantic entity
z p+: perceptions evoked by NL(p)
z p+: meaning of p; denotation of p
Propositions
z Most Swedes are tall
z Icy roads are slippery
z Speed limit is 65 mph
z Check out time is 1pm
Commands
z Keep under refrigeration
z Handle with care
9 /168 LAZ 6/9/2010
UNSHARPNESS OF CLASS
BOUNDARIES=FUZZINESS
cointension
z precisiand = model of meaning
z extension= name-based meaning
z intension = attribute-based meaning
z cointension = qualitative measure of proximity of
meanings
= qualitative measure of proximity of
the model (precisiand) and the object
of modeling (precisiend)
11 /168 LAZ 6/9/2010
CONTINUED
z A desideratum of precisiation is high
cointension.
z Informally, a precisiand is cointensive if
it is associated with a high cointension
in relation to the precisiend.
z Cointension principle: Achievement of
cointensive precisiation generally
requires that if the precisiend is fuzzy
so must be the precisiand.
0 30 130 180 0 54 0 5
Speed Throttle Shift
Control Rules:
1. If (speed is low) and (shift is high) then (-3)
2. If (speed is high) and (shift is low) then (+3)
3. If (throt is low) and (speed is high) then (+3)
4. If (throt is low) and (speed is low) then (+1)
*5 *.4
globule
0 0 0 0
graduation
extensional intensional
Name-based Attribute-based
z Generalized-constraint-based
semantics is a generalization of truth-
conditional and possible-world
semantics. In the following, precisiation
of propositions through the use of
generalized-constraint-based
semantics is discussed in greater
detail.
ED= POPULATION.SWEDES[Name;
Height]+
TALL[Height; µ]+
MOST[Proportion; µ]
1 N
t = μ most ( ∑ μ tall ( hi ))
N i =1
z Definition 2. A proposition, p, in a
natural language may be interpreted as
an implicit constraint. In p, an implicit
variable, X, is constrained by an implicit
relation, R. R plays the role of a
restriction on the values which X is
allowed to take. Generally, R is a
predicate.
X R (fuzzy set)
p: Most of Robert’s friends are rich.
p Proportion (rich.friends.Robert/
friends. Robert)
is Most X
91 /168
R (fuzzy set) LAZ 6/9/2010
CONTINUED
z In the following, Definition 2 will be
employed in preference to Definition 1.
GC(X): X isr R
X is R Poss(X=u)= µR(u)
98 /168 LAZ 6/9/2010
CONTINUED
z In representation of p as a generalized
constraint, p: X isr R, there are two
important points that have to be noted.
First, X need not be a scalar variable. X
may be vector-valued or, more
generally, have the structure of a
semantic network.
z In generalized-constraint-based
semantics, the information base, IB(p),
is represented as a collection of
relations, with the names of relations
drawn, but not exclusively, from the
constituents of p.
ED=POPULATION[Name;
µHeavy.Smoker; µNon-smoker; µLung.Cancer] +
MUCH.HIGHER[Proportion 1;
Proportion 2; µ]
Correspondingly, R is identified as
most.
Digression.
In fuzzy logic, proportion is defined as
a relative ΣCount. (Zadeh 1983) More
specifically, if A and B are fuzzy sets in
U, U={u1, …, un}, the
ΣCount(cardinality) of A is defined as:
ΣCount ( A ) = Σ i μ A ( ui )
133 /168 LAZ 6/9/2010
CONTINUED
ΣCount ( A I B )
ΣCount ( B / A ) =
ΣCount ( A )
Σ i μ A ( ui ) ∧ μ B ( ui )
=
Σ i μ A ( ui )
where I =intersection and ∧ =min
134 /168 LAZ 6/9/2010
CONTINUED
In application to the example under
consideration, assume that the height of
ith Swede, Namei, is hi and that the grade
of membership of hi in tall is µtall(hi), i=1,
…, N. X may be expressed as:
1
X = ( Σi μ tall ( hi ))
N
Step 3. Construction of ED.
The needed information is contained in
the explanatory database, ED, where LAZ 6/9/2010
135 /168
CONTINUED
ED= POPULATION.SWEDES[Name;
Height]+
TALL[Height; µ]+
MOST[Proportion; µ]
CF*p=X* is R*
where
1
X * = ( Σi μ tall ( hi ))
N
R* = MOST[Proportion; µ]
137 /168 LAZ 6/9/2010
CONTINUED
Step 5. The truth-value of p, t(p, ED), is
the degree to which the constraint in
Step 4 is satisfied. More concretely,
1
t ( p , ED ) = μ most ( Σi μ tall ( hi ))
N
ED=POPULATION[Name; µObese;µOvereat]
+ MOST[Proportion; µ]
SMALL.CITY[City.Name; µ] +
NEAR.SF.CITY[City.Name; µ]
X*=City.NameLOCATION.RESIDENCE[Name
= Carol; City.Name]
R*=SMALL.CITY[City.Name; µ] ∩
NEAR.SF.CITY[City.Name; µ]
City.NameLOCATION.RESIDENCE[Name=
X*=
Carol; City.Name] is
SMALL.CITY[City.Name; µ] ∩
R*=
NEAR.SF.CITY[City.Name; µ]
R*=Income1;µSIGNIFICANTLY.HIGHER[Inco
me1; Income2=IJP; µ]
Step 3. Construction of ED
POPULATION.SWEDES[Name;
Height] +
POPULATION.ITALIANS[Name;
Height] +
MUCH.TALLER[Height1; Height2; µ]
+ MOST[Proportion; µ]
1
qi = μ MOST [Pr oportion = Σ j aij ; μ ]
n
Compute the proportion, X*, of Swedes
who are much taller than most Italians
1
X * = Σi qi
m
157 /168 LAZ 6/9/2010
CONTINUED
Precisiation of R
R*= MOST[Proportion; µ]
The precisiated canonical form may be
expressed as CF*(p)= X is R, where
1
X * = Σi q i
m
R*= MOST[Proportion; µ]
Additional examples may be found in
cited papers.
158 /168 LAZ 6/9/2010
SUMMATION
z Precisiation of meaning is a
prerequisite to achievement of higher
levels of mechanization of natural
language understanding. Precisiation
of meaning plays a particularly
important role in communication
between humans and machines.
z A generalized constraint is an
expression of the form, X isr R, in
which X is the constrained variable, R
is the constraining relation and r is an
indexical variable which defines the
way in which R constrains X. The
primary constraints are possibilistic
(r=blank); probabilistic (r=p); and
veristic (r=v).
z Representation of p as a generalized
constraint is the centerpiece of
generalized-constraint-based
semantics, GCS. Through precisiation
of meaning, GCS opens the door to a
wide-ranging enlargement of the role of
natural languages in scientific theories.