Professional Documents
Culture Documents
1. Hyponymy: set membership of words; e.g., couch of furniture (a couch is inluded in the
set of furniture).
2. Entailment: invarianty true implication of a sentence because the first contains the
second;e.g., Bill ran entails bill did something (running includes the notion of action and
so an expression with run necessarily implies an prby bill is necessarily true and vice
versa . synonyms are words that can be subtituted.for each other an expression
whithout affecting the truth value ; i bought a'couch has the same truth conditions as I
bougt a sofa.
Furthermore, we can refine this concern for truth conditions to determine more
subtle types of likness, difference, inclution of meaning. Lexical semanticist (those who
study word meaning) typically identify two mayor kinds of antonymy- ungradable and
gradable antonyms- determined by their truth conditions. ungradable antonyms are
words whose meanings mutually exclude each other, lexical contradictions,denotations
whit no middle ground.A standard test for such antonyms is as follows: the truth of one
requires the falsity of the other(x-Y and Y- X) and the falsity of onw requires the truth of
the other ( -X-Y and -Y-X) By this test, alive and dead are ungradable antonyms: if
something is alive, it is not dead', if something is dead , it is not alive: if something is not
alive, it is dead ", if something is not dead it is alive ungradable antonyms are
sometimes called coplementaries because the complite each other in excluding the
middle ground.
if we slighly lossen the formal test for ungradable antonyms, we get gradable
antonyms(or contraries) those that have a middle ground. for gradables, only the first
part of the formal test for ungradables holds; the truth of one requires the falcity of the
others. Hot and cold meet this simpler test: if something is Hot, it is not cold ", if
something cold, it is not Hot note, however that if something is not hot ,then it not
neccessarily cold, but, perhaps, warm. Hence, hot and cold are antonyms whit a middle
ground, determined by the truth of the imperences that can be made from expression
containing them ( she cruse 1986 for fulled discussion ).
this final example return us to our original reasonession with do something).
Formal semantics: we can anumerate that influences tha hold across statement trouhgt
analysis of the content of the expression those those statments take. Semantics relations
are all kinds of infernce and are calculated on the basis of the semantics repretentations
of the expressions involved : long expression entails another because the semantics
reprecentation of one is a necesary semantic consequence of the semantic
representation of another . Formal semantics determines the nature or reprecentations
proper, so inferences are made with respect to a model for interpretation and can be
defined in relation. thereto.
Two probllem with Formal semnaties
We have seen how a formal aproach to meaning elusidates the nature and content of
semantic representation. in partikular, wehave seen how the idea of a model as and
abstract "the notatinal space" clarifies basic quetion of meaning. In spite of this great
advantage, there are certain problems inherentin formal sematics. to say that a system is
formal is to say that it is categorical and contentless. It these two features tha ultimately
force apart formal semantics and linguistic semantic.
Before we discuss these two features in relation to linguistic meaning, we should be
aware of an important caveat. Formal does mean 'formalism.' A popular method of
debuking logic and of showing the diva variety ergence of formal and non formal
semantic is to point out that the notation of formal semantic is insentive distinctions
made in ordinary language. For example, it is easy to sow that predicate logi, the
notation for quantifiers excludes of quantifiers that are otherwise found in ordinary
languange.predicate logic has two quantifiers: the universal quantifier (' every/all '), and
the exitential quantifier (' at least one/some ').The is no way to refresent the quantifiers
many, a couple, several, and so on, quantifiers found commonly in everyday speech. Ist
is sometimes thought that predicate logic and ordinary languange are simply
imcompatible becouse of this absence. This is wrong. We could certainly develop a form
of predicate logic that has symbols for these additipnal quantifiers. All we have to do is
assign them a notation: M for many, S for several, and so on. The trick is not the
notation. The formalism is arbitrary, and any phenomenon can
can be refresented. The inadequacy of the notation of formal systems is a trivial
objection. The heart of the matter lies in what makes the system formal, not in
whatmakes the formalism.
Categoricalness
A formal system is categorical: The logical objects and operatins that populate a formal
system have descrete boundaries. There is no overlap, and one object categorically
axcludes another. For example, ~, logical negation, categorically reverses the truth value
of whatever it attachhes to. There is no sense in propositional logic of more or less
negation, or a gradience of falsity. The same may be said for the quantifiers in predicate
logic. The universal quantifier A, for example, stands categorically for the totality of a set;
the existential quantifier E categorically defines ' some' or ' at least one.' There is no
gradience of quantification in predicate logic, just as there is no gradience of truth value
for the negative operator in proposition. al logic.
But if we look at ordinary language, We find that it is full of gradient phenomena,
more technically known as fuzziness (see Lakoff 1972 for the classic study; see also
Lakoff 1987, 1988; Jackendo of 1983: chap.7) the insight behing fuzziness indicates that
categories have vague boundaries and are internally organizid from central focal value,
the frototype (Rosch 1973, 1975), to less focal instances and fringe values. a category
may be said to have degree of inclusiviness rather than strict criteria that ultimately
draw a conceptual line beetween phenomena
admitted and chose excluded.
The need for gradience is especially pertinent inthe case of semantic categories, for
in example, an account of natural language negation. Negation in formal semantries is
categorical, but natural language requires degrees of negation.
Two reliable tests of the presence of negation in English are: (1) coocurrence with
any asopposed to some and (2) the eddition of a positive tag question (see horn;
Klima1964; J. payne 1985; see also chap. 9) The following negative sentence takes both:
Neither seldom nor rarerly is overtly negative (there is no not, as in [6], but they trigger
any and so must have a negative in their semantic representation, sometime like 'not
freqeuntly.' Unfortunately for the negative analysis, they also trigger both kinds of tag
questnos. so, even though seldom and rarerly ere negative in some sense, they are
totally negativthis picture is further complicated by the following:
8a. Rather than take any bribes, the senator resigned.
b. Before the senator took any bribes, he/she covered his/her tracks.
c. did the senator take any bribes?
These sentences allow any, but they are allclearly positive and so there is no obvious
need for a negative in their semantic representations. it might be argued. however, that
rather is an implicit negative, expressing something like 'so as not to.' Before might also
be argued to be negative, 'unrealized action.' A similar explanation also accounts for the
acceptability of any with questions. Imformation questions are desaigned to indicate
'uncertainly,' strictly speakin another form of negation.
Example ( 6-8 ) illustrate the gradient status of negaton. There is focal negation,
or denial, signalled by not, which allows both reliable tests of negation. There are also
markers of attenuated negation, like seldom and rarely: these take one of the test clearly
and arguably one of the others . there are finally, markers of uncertainly and
unrealization, a kind of a weak negation : not denial, but negation on the level of
expectation. these also trigger a negative test.
it seem clear, then, that ordinary language admits negation by degree, in such a
case , the categoricalness in language, in this case, the formal nature of logical analysis
prevents accurate description, and linguistic semantics, because it must deal with all
there forms of negation , has to be nonformal on this count,e at least