Z problemów współczesnej semantyki kwantytatywnej

Abstrakt

The use of computers in contemporary computer linguistics not only made it possible to work out statistically large bodies of texts or lexicons, but also permitted tox shift from the stage of material research to theoretical investigations. Zipf᾽s works gave momentum to these researches. He worked on the relation between frequency and significance, frequency and the length of expressions and etc. All those researches were concerned with the plan of expression, i.e. the linguistic form.

Recently, there have been discovered new statistic regularities which were observed in lexical semantics. This discovery is a complete novelty in quantitative linguistics whosse research have not dealt up to now with the semantic level of a language. Kryłow᾽s discovery marked a breakthrough. He proved in an empirical manner that statistic semantic regularities are revealed exclusively thanks to the operation of the law of large numbers.

Three statistic-semantic regularities are the subject-matter of the paper. The first, defined as Kryłow᾽s law, occurs in lexicons; it is a relation between the number of meanings of a lexeme (x) and the contribution of lexemes to a dictionary (W_x) which lexemes bear x meanings. It is an inversely propotional relation: unequivocal lexemes make the largest number of entries in a dictionary, as the degree of polysemy increases the contribution of lexemes decreases. In the paper Kryłow᾽s law has been tested comparatively for given dictionaries of Polish, Russian and English languages, taking into account: a) the whole lexicon, b) nouns and verbs.

The second regularity concerns vocabulary in a text. It is a distribution of meanings in a text of a polysemic word. This distribution reveals itself while investigationg into the uses of the words which are highly polysemic, e.g. functional expressions (prepositions conjunctions). Their meaning, grouped in a sequence according to their frequency in the text under investigation, reveals particular regularity called Beothy᾽s law. The paper presents few works devoted to this problem and discusses distributions of the grammatical function of the Polish preposition "w" (in) which was investigated in a passage from a book "Droga wiodła przez Narwik" (The Route Went via Narvik) by K. Pruszyński. The way in which these functions were described in contemporary Polish dictionaries is far from being perfect, thas is why A. Weinsberg᾽s test of questions has been taken as the basis for their isolation. Employing this test, the frequencies of the functions of the grammatical preposition "w" (in) have been investigated in three types of contemporary texts (i.e. realistic novel, popularized scientific text on astronomy and causeries). It turned out that the sequence of those functions depends in fact on the kind of text we deal with.

The third problem discussed in the paper is the noun definition sequence of the type: ammonia − gas − substance, made of terms of the genus proximum type on the basis of lexical definitions. In large sets which encounter e.g. 1000 sequences, made for 1000 lexems randomised from a dictionary, there are revealed also some statistic regularities concerning the length of those sequences, number of hiperonims at particular levels of sequences etc.

The final hiperonisms in sequences turned out to be extremely interesting. Their analysis is connected with a philosphical problem of searching after semantic indefinibilia.

In the case of forming sequences for all meanings of one polysemic lexem of the type "dom" − house (1. "building" 2. "home" 3. "family" 4. "house" 5. "trade institution" 6. "the general body of affairs..." − data taken from "The Small Dictionary of the Polish Inaguage" ed. by S. Skorupka et al., Warsaw 1968) there is created the so-called family of words which illustrates the whole hierarchical structure of all hiperonims for the lexeme "house".

Investigation on the family of words was introduced by R. Hammerl on the basis of the German language. Then taking a particular final hiperonim as the starting point in a larger set of family of words which hiperonim was of the type "object" one can form on the basis of all sequences, ended with this hiperonim, the socalled conceptual family seat which grasps the hierarchical structure of all hiponims subordinate to it.

The families of words and concepts make in quantitative semantics a completely new matter for investigation which deals with the hierarchical structure of concepts grasped in the dictionaries of a given language.