COLING 82, Z Horeck~ (ed.) 
North-Holland Publishirlg Company 
© Academia, 1982 
QUANTIFICATION OF MEANING AND THE COMPUTER 
Marie T~itelov~ 
Czech Language Institute 
Czechoslovak Academy of Sciences 
Prague 
C.S.S.R. 
One of the major tasks of the quantitative se- 
mantic analysis is to disclose complex rela- 
tions of sememes in communication, i.e. on the 
basis of their associations in the frame of syn- 
tactic structures. With the aid of computer it 
is possible to prepare a corpus of language ma- 
terial giving the possibility to quantify /I./ 
semantics of syntactic functions, /2./ lexical 
meanings, /3./ meanings of morphological cate- 
gories, esp. those of parts of speech, and to 
create a new type of semantic frequency dictio- 
nary. 
The assistance of computer in present-day quantitative linguistic 
studies makes possible to quantify not only language events, but 
also their relations, not only their static features, but also 
the dynamic tendencies of language. This refers, in the first 
place, to the phenomena of lexicon where the application of quan- 
titative methods as well as the modern computer processing have 
been of a certain tradition; but the computer-aided research is 
applied on other language levels, too, e.g. in graphemics, pho- 
nemics, morphemics, and can be performed also in morphology and 
syntax. Thus, the contribution of L. UhliFov& a. collaborators, 
workers of the Dept. of Mathematical Linguistics and Phonetics of 
the Czech Language Institute, informed about the computer-aided 
quantification of patterns of simple, compound, and complex sen- 
tences and about'their close relationship to the morphological 
#rid lexical Levels. 
The results achieved in quantitative analysis stimulate us to Lake 
a further step, i.e. to quantify the semantic aspects of language 
377 
378 M. Tt/~;ITELOV.~ 
phenomena. In this stage of analysis the computer is an indispen- 
sable instrument of research. 
Let us remember in this place the thirties and fifties of this 
century when the first semantic frequency dictionaries of English 
were - manually - compiled \[I\], \[2\]. They were isolated achieve- 
ments combining quantitative and qualitative approaches to obtain 
the frequency of individual lexical meanings. 
In the Dept. of Mathematical Linguistics and Phonetics, we dispose 
with a statistical corpus of texts processed by computer the units 
of which /words/ havebeen encoded in context with regard to syn- 
tax, morphology and lexicon; the corpus provides an adequate basis 
for the quantification of the so-called context meaning, i.e. 
meaning in a certain context \[3~. 
As it is known, a discourse is determined by external and internal 
linguistic conditions, in other words, it is a function of lingu- 
istic and extralinguistic variables. The complex relations between 
objects and other events of reality are reflected in language by 
syntagmatic and paradigmatic relations; to disclose their mutual 
conditioning is one of the major aims of the semantic analysis. 
The lexicon introducing the sememes within their proper systems is 
not capable to cope with their complex relations in the communica- 
tion, and, thus, it is necessary to examine the linking of sememes 
in the frame of syntactic structure. This is a task which cannot 
be done without an assistance ~f computer, especially when the re- 
lations are to be quantified. And, at the same time, this is a 
task which ought to be performed by the quantitative linguistics 
as a first step to the quantification of meaning which is today at 
the centre of interest in linguistics, logic and other sciences. 
A prerequisite for obtaining good results in the semantic research 
is a computationally processed corpus of linguistic units /sentence 
tokens/ in which we may quantify: /1./ semantics of syntactic 
functions, /2./ lexical meanings by means of which the semantics 
of sentence elements are expressed, /3./ meanings of morphological 
categories, especially those of parts of speech, when they prove 
to be of importance for semantics in points /1./ and /2./. 
On this subject we add a few remarks. To point /1./: It is well 
QUANTIFICATION OF MEANING AND THE COMPUTER 379 
known that in spite of a considerable progress in sentence seman- 
tics in these last years ~4\], \[5\], there is a number of questions 
remaining unsolved or open. Nevertheless, we are able to quantify 
semantics on the sentence level applying the principles of L. Tes- 
nitre \[6\]. The approach consists in searching for semantic compo- 
nents corresponding to sentence members in different syntactic 
functions. This is certainly no easy task, but is possible under 
the condition that we dispose with a perfectly performed syntactic 
analysis of whole sentences /simple, compound, and complex/ and 
larger samples of running texts, as is the sampling used inthe Depto 
of Mathematical Linguistics for the mentioned corpus of texts. 
This sort of underlying language material allows us to study the 
semantics of language units even larger than a sentence - this, 
however, being a prospect for a future work. 
Remarks to point /2./: The study of lexical meanings fixed ~n ex- 
planatory dictionaries shows some interesting aspects, too. The 
analysis based on their distribution in text and in relation to the 
respective semantics of syntactic functions gives a new picture of 
the frequency distribution of different context meanings associated 
with a word, and, in some cases, even corrects their position with 
regard to morphological categories. 
Remarks to point /3./: Each part of speech, of course, "behaves" 
in a different way as far as the semantics of syntactic functions 
and the lexical meanings are concerned; sometimes, the meaning can 
be significantly influenced also by morphological categories. 
E.g. the animate and inanimate genders of masculines in Czech con- 
cern not only the semantics of the analyzed word, but also its 
"environment". So for example, the animateness of subject /e.g. 
agent/, or object /e.g. afficient/, is reflected in the verb /pre- 
dicate/ and also is connected with certain lexical meanings; simi- 
larly, the verb categories /e.g. mood and tense/ often influence 
the use of lexical meanings as well as the constitution of new mea- 
nings. 
These assertions will be now exemplified on two fundamental parts 
of speech, the verb and the noun, which in the quantitative seman- 
tic analysis manifest some specific features referring to their 
part-of-speech properties and syntactic functions. 
380 M. TF.SITELOV./~ 
E.g. the Czech verb in the predicative position represents toge 
ther with its actants the semantic nucleus of sentence. The actants 
are formally expressed by the subject, object and/or adverbials as 
free complements. After the determination of meanings correspon- 
ding to the syntactic functions of verb as a further step we ascer- 
tain which of the lexical meanings functions most frequently as 
predicate. The semantics of predicate may be also ingluenced by the 
part-of-speech appurtenance of actantsassigned tosyntacticfunctions. 
assigned. It certainly makes a difference when a subject is ex- 
pressed by a noun or a pronoun, or when it is not expressed at all. 
There are other important aspects, too, e.g. whether the function 
of subject /e.g. agent/ is expressed by an animate or an inanimate 
noun, whether it is a human being, animal or thing, whether the 
word stands in the singular or the plurals whether it is plurale 
tantum etc. In the relation of object to predicate we must take 
into account the lexical meaning of object and its morphological 
categories, especially case. 
Between the verb, subject and object there exists a relation of se- 
mantic congruence, i.e. a classificatory agreement of noun and verb 
/i.e. of the predicate and its actants/. E.g. water, wine flows..._._.,___~_~_# 
not so the table, the earth~ we chase the d~ not ~. 
This principle is rather close to the intention of verb in the 
sense of E. Pauliny \[7\]. 
With the noun in Czech we quantify the semantics corresponding 
especially to subject, object, attribute and adverbial. Compared to 
the previous remarks on the semantic quantification of verb we have 
to do for the most part with an inverse relation; as a point of de- 
parture we take now the noun in one of its syntactic functions 
/with different meanings/ and procede to the verb syntactically 
functioning as a predicate. Further steps of analysis are practi- 
cally the same as with the verb, i.e. the determination of a corres- 
ponding lexical meaning of the given noun, of its morphological ca- 
tegories and/or syntactic functions of its ='environment." With the, 
adverbial the attention is paid especially to whether it is obliga- 
tory or optional .\[5\]. 
The semantic analysis of the remaining parts of speech in sentence, 
e.g. formal words /grammar words/, such as prepositions and conjunc- 
tionss have some aspects common with verbs or nouns and, in addition, 
QUANTIFICATION OF MEANING AND THE COMPUTER 381 
some special features connected with their status. In any case, 
the quantification of their semantics in the manner explained shows 
the functioning of wobds in text in a new light. 
The quantitative Linguistics can fulfill the task presented with 
the aid of computer which creates optimum conditions for the appli- 
cation of quantitative methods in semantics. Within the quantita- 
tive microanalysis Qf individual words in text /in context/ aimed 
at an identification of context meaning it is possible to obtain 
new facts concerning the theory of semantics in general and the 
quantitative semantics in special~ the application of results may 
lead to a new type of semantic frequency dictionary. And this is 
at present, under my leadership, the principal task of the Dept. 
of Mathematical Linguistics and Phonetics, in the Czech Language 
Institute. 
382 M. TESITELOV.~ 

REFERENCES

~\] Lorge~ I. - Thorndikes E.L.t A Semantic Count of English Words. 
/New York 1938/. 

EZ\] Westt M.t A General Service List of English Words with Seman- 
tic References and a Supplementary Word-List for the Writing 
of Popular Science and Technology /London 1953/. 

~\] Hannappelt H. - Melenkt H.t Alltagssprache. Semantische Grund- 
begriffe und Analysebeispiele /M~nchen 1979/. 

\[4\] Zimekt R. t S@mantick~ v~stavba v~ty /Praha 1980/. 

\[5\] Panevov&t J-t Formy a funkce ve stavb~ ~esk~ v~ty /Praha 1980/. 

\[6\] Tesni~ret L.t Elements de syntaxe structurale /Paris 1959/. 

\[7\] Pauliny, E.t ~truktOra slovensk~ho slovesa /Bratislava 1943/. 
