COLING 82, J. Horeck.f, (ed.

)
North-Holland Publishing Company
Academia, 1982
TEST-SCORE SEMANTICS
FOR NATURAL LANGUAGES
L o t f i A. Zadeh
Computer Sci ence Di v i s i on
Uni v er s i t y of Ca l i f o r n i a
Ber kel ey, Ca l i f o r n i a
U.S.A.
Test - scor e semanti cs i s based on t he premi se t hat
al most ever yt hi ng t hat r el at es t o nat ur al l anguages
i s a mat t er of degree. Viewed from t h i s per s pec t i v e,
any semanti c e n t i t y i n a nat ur al l anguage, e . g . , a
pr edi c at e, p r e d i c a t e - mo d i f i e r , p r o p o s i t i o n , quant i -
f i e r , command, quest i on, et c. may be r epr esent ed as
a system of e l a s t i c c ons t r ai nt s on a c o l l e c t i o n of
obj ect s or der i ved obj ect s i n a uni ver se of d i s -
course. I n t h i s sense, t es t - s c or e semanti cs may be
vi ewed as a g e n e r a l i z a t i o n of t r u t h - c o n d i t i o n a l ,
pos s i bl e- wor l d and model - t heor et i c semant i cs, but
i t s expr essi ve power i s s u b s t a n t i a l l y gr eat er .
INTRODUCTION
Test - scor e semanti cs r epr esent s a break wi t h t he t r a d i t i o n a l approaches t o seman-
t i c s i n t hat i t i s based on t he premi se t hat al most ever yt hi ng t hat r el at es t o
nat ur al l anguages i s a mat t er of degree. The acceptance of t hi s premi se e n t a i l s
an abandonment of b i v a l e n t l o g i c a l systems as a basi s f or t he anal ysi s of nat ur al
l anguages and suggests t he adopt i on of f uzzy l ogi c (Zadeh ( 1975) , Bellman and
Zadeh (1977), Zadeh (1979)) as t he basi c concept ual framework f o r deal i ng wi t h
nat ur al l anguages.
In f uzzy l o g i c , as i n nat ur al l anguages, al most ever yt hi ng i s a mat t er of degree.
To put i t met aphor i c al l y , t he use of f uzTy l ogi c may be l i kened t o wr i t i n g wi t h
a spr ay- can, r at her than wi t h a b a l l - p o i n t pen. The spr ay- can, however, has an
adj ust abl e o r i f i c e , so t hat one may wr i t e , i f need be, as f i n e l y as wi t h a b a l l -
poi nt pen. Thus, a commitment t o f uzzy l ogi c does not pr ecl ude t he use of a bi va-
l ent l o g i c when i t i s appr opr i at e t o do so. In e f f e c t , such a con~i t ment merel y
pr ovi des a language t h e o r i s t wi t h a much more f l e x i b l e framework f or deal i ng wi t h
nat ur al l anguages and, e s p e c i a l l y , f o r r epr esent i ng meani ng, knowledge and st r engt h
of b e l i e f .
An aci d t e s t of t he ef f ec t i v enes s of a meani ng- r epr esent at i on system i s i t s a b i l i t y
t o pr ovi de a basi s f o r i nf er ence from premi ses expressed i n a nat ur al l anguage.
I n t h i s r egar d, an i n d i c a t i o n of t he c a p a b i l i t y of t es t - s c or e semanti cs i s pr o-
vi ded by t he f ol l owi ng exampl es, i n whi ch t he premi ses appear above t he l i n e and
t he quest i on whi ch may be answered i s st at ed below i t .
(a) Duri ng much.of t he past decade Pat earned f a r more than a l l of hi s cl ose
f r i ends put t oget her
How much di d Pat earn dur i ng t he past decade?
(b) Most t a l l men are not f a t
Many f a t men are bal d
Bi g i s t a l l and f a t
425
426 L.A. ZADEH
How many bi g men are bal d?
(c) I f X i s l ar ge then i t i s not l i k e l y t hat Y i s small
I f X i s not ver y l arge" then i t i s ver y l i k e l y t hat Y i s l ar ge
X i s not l ar ge
How l i k e l y i s i t t hat Y i s more or l ess smal l ?
In f uzzy l o g i c , t he answer t o a quest i on i s , i n gener al , a p o s s i b i l i t y d i s t r i b u t i o n
(Zadeh ( 1978) ) . For exampl e, i n t he case of (a) t he answer would be a p o s s i b i l i t y
d i s t r i b u t i o n i n t he uni ver se of r eal numbers which associ at es wi t h each number u
t he p o s s i b i l i t y , ~ ( u ) , o ~ ( u ) ~ I , t hat u coul d be t he cumul at i ve income of
Pat gi ven ( i ) t he premi se, and ( i i ) t he i nf or mat i on r esi dent i n a dat abase.
In t est - scor e semant i cs, a semanti c e n t i t y such as a pr opos i t i on, pr edi c at e, pr edi -
c at e- modi f i er , q u a n t i f i e r , q u a l i f i e r , command, quest i on, e t c . , i s represent ed as a
system of e l a s t i c const r ai nt s on a c o l l e c t i o n of obj ect s or der i ved obj ect s i n a
uni ver se of di scour se. Si mpl e examples of semanti c e n t i t i e s whose meaning can be
represent ed i n t hi s manner ar e t he f ol l owi ng:
I . Anca has a young son. ( Pr opos i t i on. )
2. When Dan i s t i r e d or t ense, he smokes a l o t . ( Condi t i onal p r o p o s i t i o n . )
3. I t i s not qui t e t r ue t hat John has ver y few cl ose f r i ends . ( Tr u t h - q u a l i f i e d
p r o p o s i t i o n . )
4. I t i s ver y l i k e l y t hat Mari e wi l l become wel l - known. ( Pr o b a b i l i t y - q u a l i f i e d
p r o p o s i t i o n . )
5. I t i s al most i mpossi bl e f o r Manuel t o be unki nd. ( P o s s i b i l i t y - q u a l i f i e d
p r o p o s i t i o n . )
6. Expensi ve car . (Fuzzy pr edi c at e. )
7. Very. ( Modi f i er )
8. Several l ar ge appl es. (Second-order f uzzy pr edi c at e. )
9. More or l e s s . ( Mo d i f i e r / F u z z i f i e r . )
I 0. Not ver y t r ue. ( Qu a l i f i e r . )
I I . Very u n l i k e l y . ( Qu a l i f i e r )
12. Much t a l l e r than most. (Fuzzy pr edi c at e. )
13. Br i ng me sever al l ar ge appl es. (Fuzzy command.)
14. /Who are Edi e' s cl ose f r i ends . ( Quest i on. )
Al t hough t est - scor e semanti cs has a much gr eat er expr essi ve power than t he meani ng-
r epr esent at i on systems based on pr edi cat e, modal and i nt ensi onal l o g i c s , i t s ex-
pressi veness i s at t ai ned at t he cost of downpl ayi ng, i f not e n t i r e l y sever i ng, t he
connect i on between synt ax and semant i cs. In p a r t i c u l a r , t he homomorphic connect i on
between synt ax and semanti cs which pl ays a cent r al r ol e i n Montague semanti cs
(Montague (1974), Part ee (1976) and a t t r i b u t e d grammars f o r programming languages
(Knuth ( 1968) ) , pl ays a much l esser r ol e i n t est - scor e semant i cs-a r ol e represent ed
i n t he main by a c o l l e c t i o n of l ocal t r a n s l a t i o n r ul es governi ng t he use of modi -
f i e r s , q u a l i f i e r s , q u a n t i f i e r s and connect i ves. In e f f e c t , t he downpl ayi ng of t he
connect i on between synt ax and semanti cs i n t est - scor e semanti cs r e f l e c t s our b e l i e f
t hat , i n t he case of nat ur al l anguages, t he connect i on i s f a r t oo complex and f a r
t oo f uzzy t o be amenable t o an el egant mat hemat i cal f or mul at i on i n t he s t y l e of
Montague semant i cs, except f or ver y smal l fragments of nat ur al languages i n which
t he connect i on can be f or mul at ed and e x p l o i t e d .
The conceptual framework of t es t - s c or e semantics i s c l os el y r el at ed t o t hat of
PRUF (Zadeh ( 1978) ) , which i s a meani ng- r epr esent at i on system i n which an essent i al
use i s made of p o s s i b l i t y t heor y (Zadeh ( 1978) ) - a t heor y which i s d i s t i n c t from
t he bi v al ent t heor i es of p o s s i b i l i t y r el at ed t o modal l ogi c and possi bl e- wor l d
semantics (Cresswel l (1973), Rescher ( 1975) ) .
In e f f e c t , t he basi c i dea under l yi ng both PRUF and t est - scor e semanti cs i s t hat
most of t he i mpr eci si on and l ack of s p e c i f i c i t y which i s i n t r i n s i c i n nat ur al l an-
guages i s p o s s i b i l i s t i c r at her than p r o b a b i l i s t i c i n nat ur e, and hence t hat possi -
TEST-SCORE SEMANTICS FOR NATURAL LANGUAGES 421
bility theory and fuzzy logic provide a more appropriate framework for dealing with
natural languages than the traditional logical systems i n which there are no gra-
dations for truth, membership and belief, and no tools for coming to grips with
vagueness, fuzziness and randomness.
In what follows, we shall sketch some of the mai n ideas underlying test-score seman-
tics and illustrate them with simple examples. A more detailed exposition and ad-
ditional examples may be found i n Zadeh (1981).
BASIC ASPECTS OF TEST-SCORE SEMANTICS
As was stated earlier, the point of departure i n test-score semantics is the assump-
tion that any semantic entity may be represented as a system of elastic constraints
on a collection of objects or derived objects i n a universe of discourse.
Assuming that each object may be characterized by one or more fuzzy relations, the
collection of objects i n a universe of discourse may be identified with a collection
of relations which constitute a fuzzv relational database or. equivalently, a state
description (Carnap (1952)). In this database, then, a derived'object would be char-
acterized by one or more fuzzy relations which are derived from other relations i n
the database by operations expressed i n an appropriate relation-manipulating lan-
guage.
In more concrete terms, let SE denote a semantic entity, e.g., the proposition
p 4 During much of the past decade Pat earned far more than
al l of his close friends put together,
whose meaning we wish to represent. To this end, we must (a) identify the con-
straints which are implicit or explicit i n SE; (b) describe the tests which must
be performed to ascertain the degree to which each constraint is satisfied; and
(c) specify the manner i n which the degrees i n question or, equivalently, the par-
ti al test scores are to be aggregated to yield an overall test score. In general,
the overall test score would be represented as a vector whose components are num-
bers i n the unit interval or, more generally, possibility/probability distributions
over this interval.
Spel l ed out i n greater detail, the process of meaning-representation i n test-score
semantics involves three distinct phases. In Phase 1, an explanatory database
frame or EDF, for short, is constructed. EDF consists of a collection of rela-
tional frames each of which specifies the name of a relation, the names of its at-
tributes and their respective domains,with the understanding that the meaning of
each relation i n EDF is known to the addressee of the meaning-representation pro-
cess. Thus, the choice of EDF is not unique and is strongly influenced by the de-
sideratum of explanatory effectiveness as wel l as by the assumption made regarding
the knowledge profile of the addressee of the meaning-representation process. For
example, i n the case of the proposition p 4 During much of the past decade Pat
--- --
earned far more than al l of his close friends put together, the EDF might consist
---7-----
-----
of the relational frames
FRIEND [Namel; Name2; p], where p is the degree.to which Name1 is a friend of
Name2; INCOME [Name; Income; Year], where Income is the income of Name i n year
Year, counting backward from the present; MUCH [Proportion; ~1, where i n is the de-
gree to which a numerical value of Proportion fits the meaning of much i n the con-
text of p; and FAR.MORE [Numberl; Number2; P]. i n which p is the degree to which
Number1 fits the description far more i n relation to NumberP. In effect, the com-
position of EDF is determined by the information that is needed for an assessment
of the compatibility of the given SE with any designated object or, more generally,
a specified state of affairs i n the universe of discourse.
In Phase 2, a test procedure is constructed which upon application to an explan-
atory database -that is, an instantiation of EDF - yields the test scores,
428 L.A. ZADEH
Tl,...
bv the
3Tn, which represent the degrees to which the elastic constraints induced
constituents of SE are satisfied. For example, i n the case of p', the test
ppocedure would yield the test scores for the constraints induced by close friend,
@, far more, etc.
--
In Phase 3, the partial test scores obtained i n Phase 2 are aggregated into an
overall test score, T, which serves as a measure of the compatibility of SE with
ED, the explanatory database. As was stated earlier, the components of T are num-
bers i n the unit interval or, more generally, possibility/probability distributions
over this interval. In particular, when the semantic entity is a proposition, p,
and the overall test score, T, is a scalar, 'I may be interpreted as the truth of
p relative to ED or, equivalently, as the possibility of ED given p. In this in-
terpretation, then, the classical truth-conditional semantics may be viewed as a
special case of test-score semantics which results when the constraints induced
by p are inelastic and the overall test score is allowed to be only pass or fai l
--
The test procedure which yields the overall test score T is interpreted as the
meaning of SE.
To illustrate the phases i n question, we shall consider a few simple examples
(a) SE 4 El l en resides i n a'small city near Oslo.
In this case, EDF is assumed to comprise the following relational
frames ( + stands for union ):
EDF 4 RESIDENCE [Name; City.Name]+
POPULATION [City.Name; Population]+
SMALL [Population; PI+
NEAR [City.Namel; City.Name2; lo]
In RESIDENCE, City.Name is the name of the city i n which Name resides; i n POPULA-
TION, Population is the number of residents i n City.Name; i n SMALL, P is the de-
gree to which a city with a population equal to the value of Population is small;
and i n NEAR, P is the degree to which City.Namel is near City.NameZ.
The test procedure which leads to the overall test score T -- and thus represents
the meaning of SE - is described below. In this procedure, Steps 1 and 2 involve
the determination of the value of an attribute given the values of other attri-
butes; Steps 3 and 4 involve the testing of constraints; and Step 5 involves an
aggregation of the partial test scores into the overall test score T.
1. Find the name of the residence of El l en:
RE! c,ty RameRESIDENCEIName=Ellen]
which means that the value of Name is set to El l en and the value of City.Name is
read, yielding RE, the residence of El l en.
2. Find the population of the residence of El l en:
n
PRE = Population
PDPULATION[City.Name=RE]
3. Test the constraint induced by SMALL:
r,guSMALLIPopulation=RE]
where ~~ denotes the resulting test score.
4. Test the constraint induced by NEAR:
T2=uNEAR[City.Name=Oslo; City.Name2=RE]
5. Aggregate ~~ and T2:
T = T1 ,. T2
where A stands for mi n i n infix position, and T is the overall test score. This
'TEST-SCORESEMANTICSFORNATURALLANGUAGES 429
mode of aggregation implies that, i n SE, the denotation of conjunction is taken to
be the Cartesian product of the denotations of the conjuncts (Zadeh (1981)).
(b) SEA During much of the past decade Pat earned far more than al l of
his close friends put together.
In this case, we shall employ the EDF described earlier, that is:
EDF 2 INCOME[Name; Year; Amount]+
FRIEND[Namel; Name2; u]+
FAR.MORE[Numberl; NumberP; n]+
MUCH[Proportion; u]
The test procedure comprises the following steps:
1. Find the fuzzy set of Pat's friends:
FP4 NamelxnFRIEND [Name2 = Pat]
i n which the left subscript Namelxu signifies that the relation FRIEND [NameP=Pat]
is projected on the domain of the attributes Name1 and u, yielding the fuzzy set of
friends of Pat.
2. Intensify FP to account for the modifier.*:
CFP 4 FP2
i n which FP* denotes the fuzzy set which results from squaring the grade of member-
ship of each component of FP. The assumption underlying this step is that the
fuzzy set of close friends of Pat may be derived from that of friends of Pat by in-
tensification.
3. Find the fuzzy multiset of incomes of close friends of Pat i n year
Year; ,
i=l,...,lO:
ICFP. 4 , AmountINCOMEIName = CFP; Year=Yeari]
In stipulating that the right-hand member be treated as a fuzzy multiset, we imply
that the identical elements should not be combined, as they would be i n the case of
a fuzzy set. With this understanding, ICFPi wi l l be of the general form
ICFPi = 6,/e1+6,/e+...+6,/em
.
where el,..., e
m
are the incomes of Name ..., Name
1'
m,
respectively, i n Year., andsl,...,,m
6 are the grades of membership of Name,;.-,
Namem i n the fuzzy se t of close friends of Pat.
4. Find the total income of close friends of Pat i n Yeari , i=l;.., 10:
TICFPi =6,el+...+ 6 e
mm
which represents a weighted arithmetic
i n Yeari.
5. Find Pat's income i n Yeari:
IPi 4AmountINCOME[Name=Pat;
6. Test the constraint induced
sum of the incomes of close friends of Pat
Year=Yeari].
by FAR.MORE:
ri$pFAR.MOREINumberl=IPi; Nurnber2= TICFPi]
7. Find the sigma-count (Zadeh (1981)) of years during which Pat's income
was far.greater than the total income of al l of his close friends:
c iq'.;
,
430 L.A. ZADEH
a. Test the constraint induced by MUCH:
TePMUCH[Proportion=C]
'16
where T represents the overall test score.
The two examples described above are intended merely to provide a rough outline
of the meaning-representation process i n test-score semantics. A more detailed ex-
position of some of the related issues may be found i n Zadeh (1978) and Zadeh
(1981)
'Research supported i n part by the NSF Grants MCS79-06543 and IST-801896.
REFERENCES AND RELATED LITERATURE
1.
2.
3.
4.
5.
6.
7.
8.
9.
Bel l man, R. E. and Zadeh, L. A., Local and Fuzzy Logics, in: Modern Uses of
Multiple-Valued Logic Epstein, G.,(ed.). Dordrecht: (D. Reidel 103-165, 1977).
Carnap, R., Meaning and Necessity (University of Chicago Press, 1952).
Cresswell, M. J., Logics and Languages (London: Methuen, 1973).
Knuth, D., Semantics of context-free languages, Mathematical Systems Theory 2
(1968) 127-145..
Lambert, K. and van Fraassen, 8. C., Meaning Relations, Possible Objects and
Possible World&Philosophical Problems i n Logic (1970) l-19.
Montague, R., Formal Philosophy, in: Selected Papers,Thomason, R.,(ed.).
New Haven: (Yale University Press, '974).
Partee, B., Montague Grammar (New York: Academic Press, 1976).
Rescher, N., Theory of Possibility
Rieger, B., Feasible fuzzy semantics,Words, Worlds and Contexts (1981) 193-
209.
10. Zadeh, L. A., Fuzzy logic and approximate reasoning, Synthese 30 (1975) 407-
428.
11. Zadeh, L. A., A theory of approximate reasoning, in: Electronics Research
Laboratory Memorandum M77/58, University of California, Berkeley, 1977. Also
Machine Intelligence 9, Hayes, J. E., Michie, M. and Kulich, L. I., (eds.).
k& York: (Wiley, 149-194, 1979).
12. Zadeh, L. A., Fuzzy sets as a basis for a theory of possibility, Fuzzy Sets
and Systems 1 (1978) 3-28.
13. Zadeh, L. A., PRUF--a meaning representation language for natural languages,
Int. J. Man-Machine Studies 10 (1978) 395-460.
14. Zadeh, L. A., Test-score semantics for natural languages and meaning-represen-
tation via PRUF, Tech. Note 247, AI Center, SRI International, Menlo Park,
CA., 1981.