Algorithm 版 (精华区)
发信人: ssos (存在与虚无), 信区: Algorithm
标 题: Structural & Morphogenesis for Language
发信站: 哈工大紫丁香 (2001年06月30日19:46:27 星期六), 站内信件
Structural & Morphogenesis for Language
Abstract This article provides a new mathematical method to process the natu
ral language (NL) and explain the catastrophe or ambiguity phenomena in lang
uages. The original idea is from R.Thom & L.A.Zadeh . To build up a complete
model for NL , we must focus on the continuity of geometry of NL , not th
e discrete analysis which has dominated the computational linguistics for mo
re than twenty years.
Section 1——HOMOLOGUAX
1.1 The double origin of language
The appearance of language in man is a response to a double need:
For a personal evolutive constraint, aiming to realize the permanence of th
e ego in a state of wakefulness.
For a social constraint, expressing the main regulating mechanisms of the s
ocial group.
The first constraint fulfills the need to virtualize predation. Man in a st
ate of wakefulness cannot pass his time like an infant of 9 months, seizin
g every object and putting it in his mouth. He has greater things to do; h
e must "think", that is , seize the things lying between exterior objects a
nd genetic forms, namely, CONCEPTS.
The second constraint expresses the need for the social body to disseminate
the information necessary for its survival, like the presence of nearby en
emy or prey. Language then works as a sense relay, allowing one individual
X to describe to another Y what he, X, is in a position to see but Y, less w
ell placed, cannot see.
The social constraint will create structures in the most unstable zone of th
e individual by an effect of interaction between hierarchical levels of orga
nization ; these structures will be on the shock waves separating sleep and
wakefulness , or genetic and spatial forms , and these shock waves will exf
oliate . A baby is equipped at birth with a stock of sensory-motor schemas ,
genetic forms that manifest themselves in the so-called ARCHAIC reflexes .
Later , toward six months , these schemas undergo a kind of melting , a g
eneralized catastrophe , coinciding with the onset of infantile babbling . T
his babbling seems like the need to expel by the process of articulation som
e of these alienating genetic forms , clearly a playful emission of forms ,
not a capture of forms . Recall that a genetic form is not fixed , but rathe
r is equipped with mechanisms o self-regulation analogous to those of a livi
ng being ; a concept occurs by superposition , by projecting the regulation
schemes of the subject onto a spatial form , an exterior image . By a geome
trical analogy , we might say that the concept forms by exfoliation from the
spatial image , where the normal coordinate along which the exfoliation ta
kes place is associated with the direction of an articulated emission whose
phonetic structure has little to do with the genetic form that gave its regu
lation to the concept. (This is Saussurian idea of the arbitrariness of sign
.) This association forms by habituation , with the emission of the sound oc
curring with the (playful or biological) use of the corresponding object .On
the other hand , the laws of combination , the syntax , of these words are
not arbitrary since they are imposed by the semantic interactions between th
e concepts , themselves defined by the regulation schemes of the subject and
thus of the concepts .
If a child spends the time between one and three years of age without other
human verbal contact ,the articulatory emission catastrophe (the babble) rap
idly degenerates into the production of a few crude sounds (the verbalizati
on of "wolf children"). The exfoliation of the semantic support space of th
e concept is inhibited by the incoherence or absence of the sounds associate
d with the object ; this results in the mental retardation or idiocy to whic
h these children are condemned . Those who have studied these wolf children
have observed their extreme reactions to some noises like the cracking of a
nut . There is little doubt that their minds are still dominated by a smal
l number of alienating forms of genetic origin . Man gets rid of these ali
enating forms by giving them a name and so neutralizing their hallucinatory
powers by fixing them on a semantic space distinct from space-time .
If man has escaped from the fascination of things through the use of langua
ge , he remains under the fascination of action incorporated as the grammar
of the language .Only the oriental philosophers have tried to withdraw the
subject of this fascination by reducing virtual action to pure contemplation
.
1.2 Syntax and archetypal morphologies
It is well known that all speech can be decomposed into elementary phrase ,
each phrase being characterized by the fact that it contains precisely one v
erb , ignoring here the difficulties (about which specialists are still deb
ating ) of the definition of the traditional grammatical categories :nouns
(substantive ), adjective , verb , preposition , and so forth . The fact th
at any text can be translated from one language to another confirms the beli
ef that these categories are almost universal .Now , given a spatiotemporal
process which we are to describe linguistically, are there any formal crit
eria relating to the intrinsic morphology of the process that enable us to
predict the decomposition into phrases?
To this end , we must start by “objectively”describing a spatiotemporal mo
rphology. In fact every linguistically described process contains prividomai
ns are the actants of the process ,the being or objects whose interactions
are described by the text .As a general rule ,each actant is a topological b
all and hence contractible ; this is the case , for example , for animate b
eings .At each moment t we contract each actant to a point , and when two
actants interact this implies that their domains come into contact in a reg
ion of catastrophe points which we also contract to a point of intersection
of the lines of the two contiguous actants. In this way we associate a graph
with every spatiotemporal process.
I then propose that the total graph of interactions describing the process c
an be covered by sets Ui such that the following conditions are met:
The partial process with support Ui is described by one phrase of the text .
The interaction subgraph contained in Ui belongs to one of the sixteen arche
typal morphologies of the following figures.
In principle each of the morphologies is generated by a construction of the
following type . In the universal unfolding of each elementary singularity ,
take a ray emanating from the organizing center and having contact of maxim
um order with the discriminant variety, and then displace this ray parallel
to itself to avoid the confusion of actants at O. Then lift this ray to the
space of internal variables ,with each actant being represented by the basin
of a minimum , to give the corresponding interaction. It may be necessary t
o cheat a bit by bending the rays in order to ensure the permanence of the a
ctants to times t = +∞. Furthermore ,certain verbs called is described by t
he archetypal morphology. The occurrence of excision morphologies , characte
ristic of sexual reproduction , is noteworthy.
This theory of the spatial origin of syntactical structures accounts for man
y facts , for example , the restriction to four actants in an elementary phr
ase and the origin of most of the cases in a language with declension : the
nominative, for the subject; the accusative , for the object ; the dative ,
with verbs having the morphology, the instrumental, with verbs having the ex
cision morphology of cutting , or ablative . The only classical case that ca
nnot be interpreted by this tableau is the genitive , which is an operation
of semantic destruction , dislocating a concept into its regulating subconce
pts in a kind of inverse embryology.
1.3 The automatisms of language
It remains to account for the palpably automatic character of the formation
of syntactical structure . To this end we suppose that each main verb type (
each archetypal morphology) is represented mentally by an oscillator , which
vanishes at the organizing center at a certain critical energy level Ec . W
hen E<Ec , it typically describes a cycle in the universal unfolding with al
most a stationary point on a sheet or a branch of the discriminant variety ;
the corresponding arc describes one of the typical sections generating the
associated archetypal morphology . For example , in a verb of the capture mo
rphology there will be a cycle C in the universal unfolding of the Riemann-H
ugoniot catastrophes with the stationary point K on the capture branch. When
E = Ec , the radius of C tends to zero and the cycle vanishes at the organi
zing center O . This gives the unstable potential V = x ^ 4 in the internal
space , and then the situation evolves toward a generic situation with the
representative point in the uv- plane following a curve close to the captur
e branch K inside the cusp . This results in the formation of two dummy acta
nts corresponding to the minim of potential , and then these dummy actants
will play an instrumental role in the capture of the concepts of the meaning
. Each dummy actact excites the concept and reduces it to its splitting int
o image + word ; finally the dummy actant unites with the word and is emitte
d as a word .
Such a process then describes the emission of s phrase of type SVO , subject
- verb -object. First the vanishing of the cycle C leads to the emission of
the verb V , and then the packet of two dummy actants liberates the subject
-object ; each of these actants interacts with the corresponding concepts ,
which it maps on the subject image axis while the dummy actant is re-emitte
d as a word (substantive) . The total morphology is that of the verb " to ta
ke "(see the following figure). Thought is then a veritable conception , put
ting form on the dummy actant arising from the death of the verb , just as t
he egg puts flesh on the spermatozoid ; thus thought is a kind of permanent
orgasm . There is a duality between thought and language : thought is a virt
ual capture of concepts with a virtual , inhibited , emission of words , a p
rocess analogous to dreaming , while in language this emission actually take
s place , as in play.
1.4 Grammar , languages, and writing
1.4.1 Grammatical categories and the typology of language
We have seen that concepts have a regulation figure , a LOGOS , analogous to
that of living beings . We might regard a grammatical category (in the trad
itional sense ) as a kind of abstract LOGOS , purified to the point that onl
y the rules of combination and interaction between such categories can be f
ormalized .From this point of view , we say that a grammatical category C is
semantically denser than a category C' if the regulation of a concept of C
involves mechanisms intervening in the regulation of C' . For example , tak
e a name of an animate being , say a cat : this cat must make use of a spect
rum of physiological activities for survival - eating , sleeping , breathing
, and so forth ; once these are satisfied , he can then indulge in less nec
essary but quite normal activities - playing , purring , and the like . Sim
ilarly each substantive has a spectrum of verbs describing the activities ne
cessary for the stability and the manifestation of the meaning of the concep
t . Since the verb is indispensable for the stability of the substantive ,
it is less dense than the noun . The adjective shares in the stable characte
r of the noun , but it is defined on a space of qualities , deeper than spa
ce -time , the support of the verb. When a category C is denser than C' , th
ere is , in general , a canonical transformation from C to C' . The inverse
transformation , however , is generally not possible.
These rules lead to the following order , in decreasing semantic density , f
or the traditional grammatical categories : noun - adjective - verb - adverb
- affixes and various grammatical auxiliaries . In the emission of a sente
nce , the meaning is analyzed and the elements are emitted in the order of
increasing density . In the model of □1.3 the density of the concept is ,
in practice , the time required by the dummy actant to reduce the concept to
the representative sign. It is much longer for a complex being like man tha
n for an inanimate object ,whose regulation is much simpler. As an example ,
the normal order of emission of a transitive phrase , subject - verb - obje
ct , would be verb - object - subject ; the object is less stable than the s
ubject , since in such s transitive process the subject survives the whole i
nteraction whereas the object may perish . (The cat eats the mouse , the mo
rphology .)The reception order , the one most favorable to th
e best reconstitution of the global meaning , is generally the opposite ord
er : subject - object - verb . Now researches on the universals of language
have shown that the pure smissive typology V-O-S is well represented . This
reflects a fundamental fact in the dynamic of communication : the act of spe
aking is initiated by the speaker , and in general he has a greater interest
in being understood than the listener has in understanding . Consequently t
he transposition of the emissive order into the receptive order is generally
carried out by the speaker , and this gives predominance to the reception t
ypology . However , the mixed emissive typology S-O-V is the most common .
An elementary sentence generally contains other ancollary elements that go t
o make up a nuclear phrase ; these are the adjuncts . The principal kinds of
adjuncts are the epithet adjectives(A-N or N-A) and genitives (G-N or N-G).
The adjective is semantically lees dense than the noun; therefore the rece
ptive typology of the epithet is N-A, the emissive A-N ; and similarly for t
he genitive :receptive N-G , emissive G-N . Since a preposition is less dens
e , an adjunct of type Pre-N is an emissive type , in harmony with the order
V-O ,while a postposition N-Post is in harmony with O-V .
The second principle governing the typology of languages is this : the free
adjuncts (those not tied to the central verb , e.g., A and G) have an invers
e typology to the verb-object nuclears .This leads to the two main types of
languages :
Emissive
Receptive
V-O
Pre
N-A
N-G
O-V
Past
A-N
G-N
English is not typical , since it has preserved from an older receptive stag
e the typology A-N for the epithet adjective and the partial type G-N in the
Saxon genitive.
1.4.2 The origin of writing
The mental reconstruction of the organizing centers of elementary fields spr
eads , by a very natural contagion , to the functional fields of the hand. T
he external variables the elementary catastrophes will be realized as spatia
l variables . The stykuzstion of an action is nothing more than a return to
the organizing center of this action . To the extent that the Riemann -Hugo
niot cusp is conceptually stabilized by the concept of division or separatio
n , the catastrophe can be realized by writing in clay with a stick the symb
ol < . Similarly the sign 人 is , like the previous one an old Chinese ideo
gram meaning to enter or penetrate , and this could well be a stylization of
the elliptic umbilici . In this way we cannot but admire the suitability of
the Chinese system of writing ; the dominant influence of the spoken word i
n the West has resulted in an alphabetical or syllabic system if writing , a
nd the expression has violently subjugated the meaning .
In conclusion ,we have seen that an analysis of the grammatical structures o
f language requires a subtle mixture of algebra , dynamics , and biology . W
ithout pretending to have a definitive answer to a problem whose difficulty
can scarcely be measured , I venture to suggest that these ideas may contain
something of interest for many specialists .
Section 2——GEOMETRY OF LANGUAGE
2.1 What does it means to Understand Language ?
The domain of linguistic structure which is concerned with the structural e
lements of language.
The domain of correspondence between linguistic structures and the world ;in
other words, what do the structural elements of language refer to ,or "mea
n".
The domain of cognitive processes which involves the structure of knowledge
and the manipulation of the items in the structure by the processor of the l
anguage (either human or computer ).
The domain of human and interaction which views language within time , rela
tive to past language use and future expectations.
In some sense , computational linguistics is the main part of artificial int
elligence. The machine must learn to “understand”natural languages and ope
rates under some rules that man determines. Thus, we have to make the natura
l language to the formal one which the machine can accept. But the logician
Richard Montague did not make difference between them ,what's more , he cl
aimed that a subset of English could be considered a formal language .He sai
d that the syntax ,semantics , and pragmatics of natural language are branch
of mathematics , not of psychology .Language is an innate capability in hum
ans ,that all humans learn a language because the structure of language is
a biological aspect of the species. So we should use mathematics to build mo
dels for all aspects of language , especially the phenomenon of ambiguity th
at is the most difficult part of computational linguistics.
There are many approaches to the study of language : lexical , morphologica
l , syntactical , semantic , and so on . Richard Montague cited Charles Morr
is' distinctions when he said , the study of language was ... partitioned in
to three branches —syntax , semantics , and pragmatics—that may be charac
terized toughly as follow. Syntax is concerned solely with relations betwee
n linguistic expressions ; semantics with relations between expressions and
the objects to which they refer ; and pragmatics with relations among expres
sions , the objects to which they refer , and the users or contexts of use o
f the expressions .
Chomsky's theory of structure of syntax has been basic for computers nowaday
s. So many linguistists believed that the formalization of language was the
main method in NLP in 1960's.But as we know , TG and ATN are convenient to g
enerate sentences but not fitful for analysis . After 1970,people began to
look for new theories to overcome those shortcomings of formal language .
These theories are
C.J.Fillmore's Case Grammar (CG)
M.R.Quilian's Semantic Network (SN)
Schank's Conceptual Dependency Theory (CDT)
Bresnan & Kaplan's Lexical Functional Grammar (LFG)
M.Kay's Functional Unification Grammar (FUG)
Gazdar's Generalized Phrase Structure Grammar(GPSG)
4,5 and 6 are called Non-transformational Grammatical Theories or Unificatio
n-Based Grammars. Each of these methods has made successfully in some part
of language. But they are not going to be mentioned here , we can find the
details in every book on COMPUTATIONAL LINGUISTICS .All these theories belon
g to the past just because they could not reveal the essence of natural lan
guage .
It seems that only the mathematical method can deal with the natural languag
e completely or almost completely . The famous mathematician Rene Thom said
in his Stabilite Structurelle et Morphogenese 1972 that any good theory abo
ut language can not be without the continuity of geometry . R.Thom proposed
the catastrophe theory in 1970's and he first realized the importance of the
applications of geometry in NLP. He also tried to give a classification of
verbs in his way . As we will see , it is difficult to find Thom's continuou
s membership function .So there is much work to do before we build up the ma
thematical model for NL .
2.2 Zadeh's Fuzzy Set Theory in Language
Suppose that M is the monoid generated by letters X1,X2,…,Xk . And we can d
efine each word W in M a real number P(W) between 0 and 1 which measures th
e creditability of W . P(W) = 0 represents that W should be got rid of and P
(W) = 1 means W is OK .In a formal system ,function P(W) is restricted by th
at the OK set P (1) can be described by finite axioms .Without loss of gener
ality , P(W) is not only the function of Xi in W but also the function of th
e contexts of W.So we can study the analytic or differential properties of P
(W) which are able to indicate that the more distance from W the less conte
xts of W effect on P(W). It is the local space-time that determines the str
ucture of the nuclear sentence which is just a one dimensional projection .
The geometry of projection p: R →R determines the type of language .
In 1965, L.A.Zadah proposed the fuzzy set theory which has developed into th
e main method dealing with the uncertain problems in our world .
In Zadah's The Concept of a Linguistic Variable and its Application to Appr
oximate Reasoning 1975 , there are many examples of the applications of his
theory in NLP including the fuzzy context-free grammar , etc .To invent the
special machine to process the natural language is nessary. Someone has pro
posed the fuzzy computer theory and other types of machines to deal with th
e problems in artificial intelligence of our real world .
Section 3——SEMANTIC MODELS
Let P be a natural process taking place in an open set W in space - time .An
y analysis of the process starts with the aim of determining in which region
s of W the morphology of the process is structurally stable . Suppose that i
t is structurally stable in an open set U W, and recall that this means th
ere is a universal model K U1 such that P ,restricted to U ,is isomorphic
to the local process P' induced from K by an embedding j : U U1 . In
this case we say that P is defined on U by a morphogenetic field or , in the
terminology of Waddington , that P has a chreod on U.
3.1 Definition of a Chreod
A chreod c in space-time R ×R is specified as follows :
By an open set U in the hyper plane t = 0 , called the initiation set of c .
As in all natural evolving processes , a point x of space - time can affect
only the events in a cone C(x) with vertex x , the successor cone of x (lik
e the light cone in relativity) , and the union of these cones C(x) over all
x in the initiation set of c is an open set W , called the zone of influenc
e of the chreod .
By an open set V contained in W and containing U in its boundary , the suppo
rt of the chreod ,written as |c| , and a (static or metabolic) morphogenet
ic field defined , up to isomorphism , on V. The set W-V is called the bifur
cation zone or umbilical zone of the chreod (see the figure below).
It can happen that the support of the chreod extends to t = ∞ , and that th
e section of the morphogenetic field at t = constant tends to a fixed limit
as t tends to infinity ; in this case the chreod is called asymptotic (see t
he figure below) .
REMARKS The idea of a chreod differs from the more general idea of the morph
ogenetic field only in the privileged role allotted to time and its orientat
ion .Irreversibility of time is justified by the fact that for natural proce
sses which depend on diffusion and are , at least partially , controlled by
parabolic equation the possibility of qualitative "retrodiction " (reconstru
cting the past from the present situation ) is much more restricted than tha
t of prediction .
3.2 A Subchreod of Chreod
A chreod J is a subchreod of a chreod K if the initiation set and thus the
support of J are contained in the support of K . The morphogenetic field on
the support of J is given by the injection |J| |K| of support .
There is an infinite number of subchreod of K whose support contains a given
point x in the support if K ; in practiced , every sufficiently small open
neighborhood of x defines such a subchreod .When x is an ordinary catastrop
he point of the morphogenetic field associated with K , there is a fundament
al system of isomorphic subchreod containing x . Recall that such a system d
oes not exist if x is an essential catastrophe point of the field . Converse
ly , any chreod is contained in a maximal chreod . If the process studied is
contained in a unique chreod , it is deterministic and structurally stable
.
3.3 The Family Tree of Chreods
Say that a chreod g is the successor of chreod K1, K2 ,… , Kr if the initia
tion set of g lies in the intersection of the umbilical zones of K1 , K2 ,
… , Kr . Generally the same chreod g can have several systems of parent ch
reods , and , conversely , the same parents can give birth to several config
urations of chreods.The complete description of a semantic model requires t
he specifying of all possible rules of succession and of all the known rules
that can diminish the inherent in determinism of the scheme , and from this
point of view it is necessary to take into account not only the abstract re
lation of succession between chreods but also their geometrical position in
the configuration .
3.4 Conditional Chreods and Levels of Organization
The examination of the corpus of an empirical morphology often shows that ce
rtain aggregates of elementary chreods occur very frequently or exhibit a hi
gh degree of stability . Often it is possible to describe constrains on the
initial condition that ,when satisfied , make such an aggregate of chreods s
table .In this case we speak of conditional chreod . Examples of such condit
ional chreods are words in linguistics and living beings in biology .
It is often possible to specify several hierarchical levels of organization
among conditional chreods .In linguistics , for example , there are the suc
cessive levels of phoneme , syllable , word ,and phrase ; while in biology t
here are supra molecular organization ,cytoplasm , cell , organ , organism ,
and population .
The decomposition of a chreod at level j + 1 in terms of chreods at the lowe
r level j is called its structure . A pair of levels j , j + 1 will be call
ed formalizable if there is an abstract morphology M (generally finite ) and
a homomorphism of the chreods at level j + 1 into the aggregates of M such
that the images of the component chreods at level j are the atoms of M (the
homomorphism being compatible with the contiguity relations ). This notion
accounts for the ideas of function in biology and grammatical category in sy
ntax . In generative grammar , for example , the structure of a sentence is
represented by a tree such as the following :
Here the abstract morphology M consists of symbols such as A (article) , N(n
oun) , and V(verb) . Syntactical analysis of a sentence is then a homomorphi
sm of the collection of words in the bottom line of the tree agreeing with
the abstract morphology M , Thus the symbols A , N ,and V represent grammat
ical functions (or ,more precisely , categories ) .
3.5 The Analysis of a Semantic Model
Let us start the analysis with the particular case of language . A first met
hod is the formal attack : neglecting completely the internal structure of e
ach chreod (here ,the meaning of each word ) , we describe the formal relat
ionships among them . To this end the basic step is to collect a stock of ex
amples sufficiently large to allow valid conclusion to be drawn about the fo
rmal relationships of succession ; this stock is the corpus of the linguists
, and from it , in principle , one can deduce the grammar which governs th
e association of chreods in the configurations found in the corpus .
The foundation of the corpus is the primordial essential task of the experim
enter, and many sciences , particularly biology , have scarcely passed this
stage . In linguistics ,thanks to our direct intuition of the meaning of wor
ds ,we have been able to produce a formal classification of words associated
with their grammatical functions and , in this way , to define formal condi
tions for a phrase to be grammatically correct , but not for it to be meanin
gful . The linguist's task is very difficult , for there is almost no conne
ction between the written or spoken structure of a word and its meaning .Th
e choice of the word corresponding to a given meaning is the result of a lo
ng historical process , a quasi-permanent generalized catastrophe . This is
due to the fact that human communication has imposed rigid constraints on th
e structure of expression.
In the natural processes that do not aim at communication and that can be as
similated in language only by metaphor (not by a useful metaphor ), we must
expect the internal structure of each chreod to be relatively transparent ;
in such a language all words will be onomatopoeic. In fact it is a good idea
to suppose , a priori , that a chreod contains nothing more than can be ded
uced by observation , that is , the catastrophe set , and to proceed to the
dynamical analysis of the chreod which is the most conservative .From this
point of view ,the significance of a chreod is nothing more than the typolog
ical structure of the catastrophes it contains and its possible dynamical i
nterpretation . This leads to the definition of the organizing center of the
chreod .
3.6 The Dynamical Analysis of the Chreods of a Static Model
We now restrict attention to chreods associated with gradient dynamics becau
se there are the only ones that are susceptible to mathematics treatment ; m
etabolic models lead , via catastrophes , to generalized catastrophes about
which we know nothing at the moment . All the chreods associated with such a
process on space-time have been described ; What about their associations ?
Some of them occur more frequently than others ; one important class of t
hese configurations is that with organizing centers of codimension greater t
han the dimension of space-time . Suppose that z is a jet that is determinat
e but , for example , of codimension five ; the universal unfolding of the s
ingularity defined by this jet is five dimensional ,and in it Maxwell's conv
ention defines a universal catastrophe K .The intersection near the origin o
f K and a transversal R will then have a configuration of chreods that ar
e structurally stable under small deformations of the global evolution of th
e process . this permits us to talk of the jet z as the organizing center of
the configuration . The local stable configurations of chreods , the gramma
r of the process , is completely specified by the topology of the bifurcatio
n set in the function space of potentials . A perturbation of the embedding
of the R in the universal unfolding R , sweeping across the origin ,
for example , will result in a modification of the stable configuration of c
hreods , and this allows us to speak of configurations being linked or dual
with respect to the organizing center z .
It may seem difficult to accept the idea that a sequence of stable transform
ations of our space-time could be directed or programmed by an organizing ce
nter consisting if an algebraic structure outside space-time itself . The im
portant point here as always , is to regard it as a language designed to aid
the intuition of the global coordination of all the partial systems control
ling these transformations .
BIBLIOGRAPHY
M.D.Harris Introduction to Natural Language Processing 1985.
R.Thom Mathematical Models of Morphogenesis 1983.
R.Thom The Double Dimension in Universal Grammar ,Les Lettres modornes , Vo
l.89,1978,pp78-96. Paris.
L.A.Zadeh The Concept of a Linguistic Variable and its Application to Appro
ximate Reasoning 1975.
--
<<社会契约论>>是一本好书,应当多读几遍
风味的肘子味道不错,我还想再吃它
※ 来源:·哈工大紫丁香 bbs.hit.edu.cn·[FROM: 202.118.230.220]
Powered by KBS BBS 2.0 (http://dev.kcn.cn)
页面执行时间:415.083毫秒