COMPUTATIONAL LINGUISTICS AND ITS ROLE IN MRCHANIZED 
OR MAN-MACHINE COGNITIVE PROBLEM SOLVING 
M. K. Chytil 
Center of Biomathematics, Czechoslovak Academy of Sciences, 
142 20 Prague 4, Vidensk~ 1083, Czechoslovakia 
In the present paper cognitive science will be conceiv- 
ed as a discipline which theoretically supports the following 
constructing of various cognitive problem solving systems 
running in a CA mode or in a man-machine mode or in the form 
of cognitive robots. The role of computational lin~stios i~ 
this context will be demonstrated and Justified. 
I. CoKnitive uroblem solving. Hence "cognition" is not 
considered here as an object of psychological analysis but 
ra~her in the sense of a man-machine cognitive process, i.e. 
of a purpose built point of view. Therefore, an analogy wlth 
industrial mass-productionwill be emphasized and the necess- 
ary theoretical questions for projecting and setting up such 
efficient mechanized or computer assisted cognitive systems 
will be studied. Such systems can be employed especially in 
scientific research since it presents a systematic form of 
activity in the field of general cognition. Following our ana- 
logy it is to say that such "factories on cognition" should 
not be identified with usual computing centers~ The consider- 
ations will be focused on a so-called co~itive, problem, It is 
a question raised for inquiry, investigation or discovery, 
which needs to be solved and where the final solution will 
present new knowledge. Non-cognitive problems are designated 
as technical problems. Their final solution consists of a de- 
sirable change in a material system. In the following we re- 
- 74 - 
strict our account to factual extramathematical cognitive 
problems the solving of which satisfies necessarily the two 
conditions: 
I. a computer is used, at leas~, partly in the solving 
I" mathematical means are included, at least, in a part of 
the process of reasoning (as an algorithm or inferring in 
a suitable calculus). 
The presence of other solving means, such as linguistic- 
al, methodological, etc. follows directly from the fact that 
only factual problems are at stake. 
2. S~nbiotio problem solvers. Cognitive problem-solving 
procedures consist of operations with concepts, expressions 
or symbols. If such a procedure has the character of an algo- 
rithm we call it a routine ~rocedure, otherwise it is termed 
a creative one. Any goal-seeking performance of such cogniti- 
ve procedures is considered to be an intelli~ent activity. We 
can by this way unambiguously say that man or machine acts 
intelligently and avoid by this way such vague expressions as, 
for instance, °be'or °it thinks'. Systems which can act in- 
telligently are called intelli~ent agents. They generally 
satisfy the following requirements- 
(i) they have the ability of performing operations 
with abstract entities, 
(ii) the~ can accept knowledge from other systems 
(agents) and then use it correctly (i.e. they understand it), 
(iii) they must be able to communicate knowledge to 
another, similar system (an agent), 
(iv) they must have a capacity for memory and be able 
to learn from their history (1). 
Intelligent agents can be used as processors for cognitive 
problem solving. Examples of intelligent agents are man, a 
computer, an animal. When connecting such agents functionally, 
but not necessarily, physically, various types of symbiotic 
intelligent agents can arise, shortly said symbionts (man- 
-computer, a group of people and computer, etc.). Let us note 
- 75 - 
that in a symbiotic connection both man and machine obtain 
certain exactly circumscribed tasks from a control unit of the 
whole system. Man is charged with tasks required by creative 
procedures ~reas machine is charged wi¢h the routine ones. 
The difference between the symbiotic and the usual interactive 
connection of man and machine consists in the fact that in the 
interactive mode man is not controlled and, therefore, he can 
act fully independently, using the computer as an efficient 
tool only. The sense of a man-machine symbiont emerges per- 
spicuously when accepting the following three assumptions: 
A. The solving of each cognitive (solvable) problem can 
be done by a creative procedure. Moreover, some of these pro- 
oedures can be algorithmized (routinized). 
B. Each routine procedure realizable by man is, in prin- 
ciple, also realizable by machine, but not conversely, i.e. 
there are some routine procedures which are realizable by 
machine only. 
C. Some of the non-routinizable procedures, as in the 
sense A, can be performed either (i) only by man, or (ii) by 
machine only, or (iii) by both man and machine. 
Obviously various basic kinds of cognitive problem sys- 
tems, particularly of symbionts can be set up. Respecting the 
assumption A we shell prefer oognitive systems where: 
(a) the solving is routinized whenever possible (ass. A) 
or necessary (ass. B) and we take into account only the mecha- 
nized processing of routine procedures, 
(b) the performance of non-routinizable procedures sa- 
tisfying ass. C(i) or C (li) will be done exclusively by man. 
Such systems have not only a broader range but they are 
not replaceable by a single computer solving system. They are, 
moreover, cheaper. The role of computational linguistics is 
quite essential in symbiotic systems because the requirements 
~O ~he .number of communications inside and outside of the sys- 
- 76 - 
tern grows rapidly. Let us note that a so-called "expert-system" 
is a vague expression for cognitive problem solving systems 
constructed onthe principles mentioned above. 
3. Computational linguistics and Sts role in the f~..t. 
ly of cognitive toolsf A mechanizing cognitive solwlng pro- 
cess runs within the following general triad: 
1. Knowledge 2. Form (method) 3. Processor 
Knowledge is partly factual (i.e. taken from the given problem 
domain), partly auxiliary. In the latter case its certain part 
is necessary for the questions arising in connection with the 
proper construction of the solving system (knowledge backed by 
computer science, mathematics, logic, methodology of science, 
psychology, systems theory, cybernetics and (computational) 
lingulstlcso The necessity of the support from the part of 
computational linguistics is given: 
- by the requirements to the communication and understandiDg 
occurring in the definition of intelligent agents, 
- by the ,assumption I., 
- by the assumption I~ and by the fact that we have only fact- 
ual problems in mind. The representing of a factual structu- 
re into a formal one presents one of the most difficult 
parts of the whole cogni$ive solving process, 
- by the preference of symbiotic cognitive s~stems. In th~s 
way the claims for various kinds and levels of linguistic 
communication and understanding grow more quickly than they 
do by the usual machine mode of processing. 
The above exemplified importance and needs of computat- 
ional linguistic tools for a symbiotic cognitive solving 
process continues obviously even under the condition that the 
use of the natural language would be, if possible, reduced, i. 
e. by replacing it by a semiformal or a formal language. Si- 
milarly computational linguistics should not avoid, in this 
connection, to cover those areas which are presented by various 
kinds of image processing and manipulation (nonverbal input- 
- 77 - 
output) runnin~ in the course of a mechanized cognitive solv- 
ing process. 
4. Cognitive science. The above described mechanized or 
symbiotic co~xitive problem solving systems, their construct- 
ion, maintenance and use present the necessary basis and rai- 
son d °etre of computational linguistics. But neither the 
computational linguistics itself nor the disciplines and means 
required by I and I ° (computers and mathematics) are suffi- 
cient for the given goals. The rest can be gathered from a set 
of disciplines called above auxiliary cognitive sciences. 
Thus cognitive science has its general methodological back- 
ground in philosophy, it studies various nonphilosophical 
methods and nonmaterial tools or srlslng in the individusl 
cognitive sciences or thanks to some interaction among them 
and from the viewpoint whether and how such tools could be 
helpful in the process of mechanized factual cognitive problem 
solving. Its object is thus a theory of mechanized cognition. 
Its method is partly mathematics, partly logic, partly philo- 
sophy. The development of cognitive science is stimulated by 
the needs of individual factual sciences such as social scien- 
ces, biology or physics or by various civic areas of knowled- 
ge. The results of cognitive science are applied in a discip- 
line called cognitive engineering which has the proper con- 
struction of the concrete mechanized cognitive problem systems 
as its duty whereas the use of such systems runs on the parti- 
cular problems in the course of the single factual disciplines. 
This approach described in (2) starts from the Bobrow's (3) 
and Collin's (4) the original determination of cognitive 
science, being, moreover, consistent with the PedoseJev's 
appeal (5)'. It differs in two essential points: firstly in the 
main pragmatical aim of cognitive science and secondly in its 
approach to mathematics. Let us emphasize that the whole act- 
ivity called usually (computer assisted) applications of 
mathematics, makes, according to our approach, a certain part 
-78- 
of cognitive engineering. By this way the assertion of comput- 
ational linguistics is, at least, more conspicious. 

References

(I) Chytil, ~. K., Mathematical methods as cognitive problem- 
solvers, Kybernetes 9 (1980), 197-205. 

(2) Chytil, M.K., Towards cognitive science and cognitive 
engineering, Teorie rozvoJe v~dy 4 (1980), 101. 

(3) Bobrow, D.G., Preface to Representing and Understanding, 
ed. by D.G.Bobrowand A.Collins, Academic Press, Inc., 
New York, San Francisco, London, 1975, pp. IX-XII+ 

(4) Collins, A., Why Cognitive Science, Cognitive Science I 
(1977), 1-2. 

(5) Pedoseev, P.N., PilosofiJa i integraciJa znaniJa, Voprosy 
filosofii 7 (1978), 16-30. 
