„Dabartinės rašomosios lietuvių kalbos dažninis žodynas“ ir jo bazė

ACTA LINGUISTICA LITHUANICA XLVI (2002), 19 – 37 A Frequency Dictionary of Contemporary Written Lithuanian and Its Database LAIMA GRUMADIENĖ Institute of the Lithuanian Language There are few electronic databases for Lithuanian. The largest are the Frequency Dictionary of Modern Written Lithuanian, compiled by V. Žilinskienė and L. Grumadienė, and the Lithuanian Language Corpus of Kaunas University. The former is presented in detail here: the principles of its compilation and its possible uses are discussed, and the hitherto published work in statistical linguistics based on it is briefly characterised. The main problems in connection with the databases for modern Lithuanian are: (1) their insufficient differentiation (they reflect only the written, not the spoken language), (2) their not being optimally used, as the Lithuanian scholarly community lacks the preparation needed for working with such tools, and (3) the fact that the efforts of scholars compiling them receive too little public recognition and are not valued as original scholarly achievements. 1. THE AGING AND “IRRELEVANCE” OF ELECTRONIC DATABASES Lithuanian linguistics is at this very moment like a tiger frozen in mid-leap: outstanding works that emerged largely from the impetus of structuralism have already been written, yet the air is thick with anticipation for new works no less significant. This requires not only a different approach to linguistic phenomena, but also new, untested methods and working tools. In this age of information technology, new tools for linguists are often associated with electronic databases, and these, in turn, with linguostatistical publications. It would seem self-evident that the two are different from each other: electronic databases are dynamic and can be used independently, whereas publications are static—they are merely collections of data presented at the authors’ discretion. Yet in the past, and sometimes even today, such databases and publications were or are conflated in Lithuania. This is illustrated by the history of reverse (also called inversed) dictionaries. The first Lithuanian reverse dictionary, compiled by the American David F. Robinson (1976) on the basis of the 1954 edition of DLKŽ, covered about 45,000 words; it may not even have had its own electronic database, at least the author makes no mention of one. Moreover, an electronic database of this kind would have been inaccessible in Lithuania at the time, since few people had even leafed through the dictionary itself, as it had been published in a small print run in America. Juozas Korsakas of Šiauliai (1991) compiled, using a selection principle, the Inversive Dictionary of the Lithuanian Language 20 | LAIMA GRUMADIENĖ dictionary, which covered almost 22,000 words. It too was based on DLKŽ, but on the 1972 edition. The work was carried out using an electronic database compiled by the author, but the latter is not publicly available (probably for technical reasons), and thus can be accessed only in printed-book form. Vida Žilinskienė’s Reverse Dictionary of Contemporary Lithuanian (1995) is not simply a reprint of DLKŽ, since the words are presented in reverse order and classified by part of speech, and a statistical analysis is included. Today it is essentially available only in printed form as well, although an electronic version also exists. It has seen relatively little use (by the working group led by E. Jakaitienė while compiling a Lithuanian–Norwegian dictionary, and for Daiva Murmulaitė’s articles (1998, 1999, etc.)). Incidentally, the latter dictionary and analysis were prepared using the computerized version of DLKŽ (1993) and its statistical analysis, and not, as Rūta Marcinkevičienė (2000a: 16) states in her review, the database of the Frequency Dictionary of Contemporary Written Lithuanian (Grumadienė, Žilinskienė 1997; 1998). A joint effort by scholars from the Institute of Mathematics and Informatics and the Institute of the Lithuanian Language produced an electronic version of DLKŽ (1993 edition), which was printed in a small run (1995), since there was simply no right to distribute publications electronically. The Institute of the Lithuanian Language has now taken up the computerization of DLKŽ again, this time the 2000 edition (led by Stasys Keinys), and has received financial support from the Lithuanian Science Foundation. As is well known, the 1993 and 2000 editions of DLKŽ differ very little, so it is worth noting that it was decided not to use the earlier electronic version solely because it had been prepared in the already morally outdated DOS environment and FOX-PRO database: computer technicians were once again set to scan the electronic text already available, which required only minor corrections, thereby increasing labor and expenses. Might the statistical analysis even be carried out all over again? But surely, just like the computerized version of the 1993 edition of DLKŽ, many electronic databases become morally outdated, so it is clear that their maintenance must be taken seriously. Could the same fate that befell the electronic version of DLKŽ (1993) also await the Terminology Bank being prepared through the joint efforts of the institutes mentioned above? Thus, linguostatistical publications are still valued far more highly than costly electronic databases, which offer many opportunities for innovative work. Lithuanian studies scholars should agree on standards for electronic databases that use special characters, for example, those containing old texts or transcribed dialect texts, since even now articles containing examples of this type have to be retyped if, for one reason or another, it is decided to submit them not, say, to publications of the Institute of the Lithuanian Language, but to one of the editorial offices of Vilnius University publications. The dialect dictionaries and dialect texts published today, although in electronic form, are merely signs of the modernization of the printing industry, since it is impossible to use their texts as an electronic database. Thus, without agreement among Lithuanian studies scholars on a common encoding system (e.g., UNICODE or another), public use of electronic data banks is greatly impeded, not to mention the fact that the use of electronic technologies and programs to analyze them becomes severely limited and sometimes impossible. Do we really have so many of them that we can no longer manage them? The problem is that neither linguists nor the computing specialists who assist them have yet properly understood that databases must be dynamic: working with them must allow the use of the many new programs and technologies that continually emerge, and they must be capable of conversion to a higher computer version. To date, most electronic databases created in Lithuania have been designed as if they were the same old slips of paper with DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 21 language data, but merely collected using a computer keyboard. Before long, these data may be comparable to cuneiform carved in stone: on the screens of next-generation computers, only “flags” and “squares” will be visible, and future linguists may undertake to decipher them, much as one reads runes. Discussions on this issue have been held before, but to no avail; still, apparently, we should not lose hope—it is worth inviting people to discuss and reach agreement once again. Incidentally, attention should be drawn to the terms for electronic formats: DATABASE, BANK, TEXT CORPUS, ARCHIVE. The most fashionable of these is TEXT CORPUS. It is defined in various ways, but all definitions share the understanding that a TEXT CORPUS is a collection of texts in electronic form; consequently, almost everything previously called a database has often come to be called a corpus (after all, the size of individual texts is not regulated). The term BANK is most often associated with terminology: banks of terms are generally not called corpora. Nowadays, the term ARCHIVE is increasingly associated with the time dimension: a collection of old electronic data may be called an archive or a database. Thus, the most neutral term remains ELECTRONIC DATABASE, which is why it is used most often here (cf. Kennedy 1998: 3 ff.). 2. ELECTRONIC DATABASES OF CONTEMPORARY LITHUANIAN AND LINGUISTIC-STATISTICAL PUBLICATIONS There is no shortage of electronic databases of contemporary Lithuanian. Linguistic-statistical publications are correspondingly few, and linguistic-statistical methods are used only somewhat more often. Phonetics, and the phonology based on its findings, stand out as the sole exception. Perhaps this is due to the nature of phonetics itself—that is, because the same units, sounds, recur frequently, conclusions based on statistics almost suggest themselves—or perhaps it is due to subjective factors, namely the emergence of researchers with a new way of thinking: above all, Aleksas Girdenis and his students, as well as Antanas Pakerys. In any case, both factors brought about a leap in Lithuanian phonetics and phonology and even a certain separation from other branches of linguistics. It is true that phonetics itself is an intermediary field between the exact sciences and linguistics. The branch of linguistics next in terms of the frequency of recurring units is morphology. It would therefore be logical to assume that this is the first field in which a need should arise to work with both linguistic-statistical data and electronic databases. So far, this has not happened, although such studies and tools already exist. The first linguistic-statistical study of lexicon and morphology in Lithuania was prepared by Vida Žilinskienė (1990), using an ESM. Her Frequency Dictionary of the Lithuanian Language is based on data from one functional style—journalism—and analyzes 300,000 word occurrences (including both occurrences of the same lexemes and of different ones; there were 18,776 distinct forms, after 11,956 proper nouns were excluded from further analysis in accordance with the generally accepted rules for compiling frequency dictionaries). All continuous texts, divided into samples of 1,000 word occurrences, were morphologically analyzed and accentuated, then punched onto cards and loaded into the now morally outdated and no longer used ESM “Minsk-22.” The electronic data bank exists, but it would now be very difficult to use. Regrettably, the linguistic-statistical analysis of journalistic texts—a variety of contemporary Lithuanian—has still not been properly put to either scholarly or practical use, especially in writing textbooks and educational dictionaries, whether from a lexical, morphological, or accentological perspective. One of the studies prepared using electronic tools (with participation in the 22 | LAIMA GRUMADIENĖ international CHILD program) and devoted to synchronic morphology is exemplified by Ineta Savickienė’s (1999) doctoral dissertation on the morphology of children’s language. Incidentally, it should be mentioned that studies drawing on linguistic-statistical data in other levels of language, in stylistics, and, to some extent, in syntax—where units occur much less frequently—had already begun to appear several decades ago, especially those by Audronė Bitinienė (1983, 1997, 1998, etc.). There have also been a few attempts to study contemporary lexicon using linguistic-statistical data. Recently, such studies have most often been conducted using the database of Vytautas Magnus University’s Corpus of Contemporary Lithuanian (approximately 60,000,000 word occurrences), for example, Rūta Marcinkevičienė’s (2000a; 2000b) studies of collocations, Jurgita Mikelionienė’s (2000) doctoral dissertation on neologisms, and others. An increasing number of lexical studies are being prepared on the basis of small electronic databases, sometimes on quite specific topics—for example, Asta Ryklienė’s (2000) doctoral dissertation on the lexicon used by internet users and her related work. Overall, however, there are few such studies, and students have been only minimally involved in this process (with the possible exception of Vytautas Magnus University). A group led by Antanas Pakerys at Vilnius Pedagogical University is completing an electronic database intended for lexical research; perhaps there will then be more student work of this kind here as well. This article is limited to describing the tools needed for research into contemporary written Lithuanian; it therefore does not discuss the already burgeoning series of studies of old writings, which draw on the Lithuanian Language Institute’s own electronic database. The emergence of electronic databases of their own will undoubtedly give a new impetus to Lithuanian dialectology: the Lithuanian Language Institute is preparing to establish a Dialect Centre, where dialect recordings and transcribed texts deciphered from them would be prepared in electronic form. Work has also begun at the same institute on creating electronic databases of onomastics and toponymy. The electronic database now being prepared for the great Dictionary of the Lithuanian Language should mark a genuine turning point in Lithuanian lexicography, and much is also expected from the electronic database of the Dictionary of Standard Lithuanian. Faculty members at Vilnius University, led by Evalda Jakaitienė, are participating in one of the European Union projects and preparing an electronic so-called universal list of Lithuanian words, which is intended to provide the basis for a new series of bilingual dictionaries. To some extent, this seems to compete with the editorial project for the Dictionary of Standard Lithuanian. With most of the work carried out by the State Commission of the Lithuanian Language, and prompted by the European Union, electronic so-called Terminology or Foreign-Word Portfolios are being prepared. These are lists of combinations of terms and words from various fields that are both recent and very frequently used; the European Union distributes them to all its member and candidate states so that their languages can adapt and prepare for the coming wave of translations, as well as for machine translation. A group of researchers at Vilnius University, led by Meilutė Ramonienė, has begun preparing electronic databases of contemporary spoken language, as in Poland, on the basis of recordings of radio and television speech. At the same university, preparations are under way to compile an electronic database of Slavicisms (led by Vytautas Kardelis), which is to include both audio recordings and examples selected and transcribed from them. Experience shows that results can be expected only after one has learned to work with data of this kind and acquired the skills to use the new possibilities. Laboratories are needed for this purpose, where people can learn to do so, preferably from an early age. The very nature of electronic databases implies openness; once organized, they should be accessible to everyone who needs them. The largest collections of Lithuanian-language data are held at the Lithuanian Language Insti- DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 23 therefore, as it modernizes, it should become the largest center for developing electronic Lithuanian-language databases, a source for their public use, and a hub attracting and disseminating work and ideas of this kind—perhaps the Lituanistics Center being established within the database of the Institute of the Lithuanian Language will take on this role? Those electronic databases that make it onto the internet will become even more accessible. Incidentally, when access to the internet becomes possible, the place where one uses a computer ceases to matter; what matters is only what has made it onto the internet. But these are plans for the future. To date, the publication of linguostatistical works in electronic form has not been formally authorized in Lithuania, and the issue of copyright has not been resolved; consequently, printed versions of such works remain relevant. Electronic databases and linguostatistical publications, however, differ fundamentally: the latter are merely the tip of the iceberg. What follows is a discussion of a specific issue: the relationship between electronic databases of contemporary written Lithuanian and frequency dictionaries. The latter cannot exist without a database, but the possibilities offered by databases are by no means limited to publishing dictionaries. 3. ELECTRONIC DATABASES OF CONTEMPORARY WRITTEN LITHUANIAN AND FREQUENCY DICTIONARIESl Electronic language databases are distinguished from other kinds of language-data collections, naturally, first and foremost by their electronic form and the specific features of the computer technologies associated with it. The advent of electronic databases opened up entirely new possibilities for language research, but they should not be associated solely with the emergence of modern electronic media. The fact that essentially analogous results—above all, frequency characteristics of grammatical forms and lexemes—can be obtained by different means shows that there is considerable common ground between modern electronic databases and their much older counterparts. In France, the Larousse publishing house had already begun using a machine-based dictionary card index (for older-generation electronic computing machines, ESM) to compile dictionaries in 1956 (Rozencveig 1986: 82). The question of how modern electronic language databases should be compiled arose, on the one hand, because the advent of new technologies made it possible to handle large quantities of information, and the amount of text used to illustrate a word (that is, its context) began to grow. This opened up new possibilities for investigating the presumptions of a word, and later of larger linguistic units, and thus syntax and semantics. There was growing enthusiasm for ever larger stretches of text, eventually extending even to complete, finite texts, until the costs of acquiring the rights to copyrighted works put a stop to this. On the other hand, the advent of computers was probably comparable in impact to Gutenberg’s invention, and, most importantly, their unprecedentedly rapid spread meant that as human connections at a distance with both speakers of the same language and speakers of other languages grew geometrically, an urgent social demand arose for obtaining diverse statistical information related to language, in order to automate as many communication processes as possible: language teaching, translation, recognition, information compression, transmission, and so on. Lithuanian linguistics did not remain outside this general process, although, naturally, it lagged considerably behind. Throughout the twentieth century, especially in its second half, the world seemed to be preparing for a new leap forward. First and foremost, these were frequency dictionaries of languages. The first appeared as early as the late nineteenth and early twentieth centuries. 24 | LAIMA GRUMADIENĖ at the end of the nineteenth and beginning of the twentieth century. The very first was a German-language dictionary, compiled by the stenographer F. Kaeding (1898) on the basis of 11,000,000 word occurrences, using a card index. Much later, G. Vander Beke (1929) published a French dictionary of this kind. H. Eaton (1934) compiled a comparative dictionary of the first thousand most frequent words in four languages—English, French, German, and Spanish. The first frequency dictionary of English was compiled by E. Thorndike and L. Lorge (1944), and the first of Spanish by V. Garcia Hoz (1953). The first frequency dictionary of Russian was compiled in America by H. Josselson (1953), while N. P. Vacar (1966), also in America, published a frequency dictionary of spoken Russian of the Soviet period, although in essence it was compiled solely on the basis of the language of dramatic works. E. Šteinfeldt (Štejnfel’dt 1963) published a frequency dictionary of Pushkin’s language in Tallinn. Based on a one-million-word corpus consisting half of fiction and half of so-called informational prose (journalism, administrative documents, and scholarly literature), a frequency dictionary of Russian was compiled in Uppsala by L. Lönngren (1993). The first frequency dictionaries not only of Russian, but also of Romanian and Italian, were likewise compiled in America by A. Juilland, P. M. Edwards, and I. Juilland (1965), and by A. Juilland and V. Traversa (1973) (incidentally, A. Juilland and E. Ch. Rodrigues (1964) had already compiled another frequency dictionary of Spanish there, followed later by one compiled by J. R. Alameda and F. Cuetos (1995)). A whole series of frequency dictionaries was based on the Prague Linguistic Circle’s conception of functional language styles, as well as Boris Larin’s understanding of the standard language. The first, as was fitting, was Czech: first dictionaries of individual functional styles, and later a general one, written by a team of authors led by Marie Těsitelová (J. Jelínek, J. V. Bečka, et al.; see Těsitelová 1983). The Latvians, like the Czechs, were quick to act, first compiling dictionaries of individual functional styles and later a general dictionary in two volumes—by T. Jakubaite, D. Kristovska, D. Guļevska, V. Ozola, R. Prūse, A. Rubine, and N. Sika (1966–1976). V. A. Agrajevas, V. V. Borodinas, V. M. Muratova, and E. V. Tisenko, under the direction of L. N. Zasorina (1977), used the same methodology to compile a frequency dictionary of contemporary Russian based on one million word occurrences from four functional styles. A frequency dictionary of Belarusian based on only one functional style—fiction—was compiled by N. S. Mažeika and A. J. Suprun (Mažeika, Suprun 1976). The frequency dictionary of Bulgarian, too, is based only on the journalistic style (Todorova, Pančovska 2001). The first Lithuanian frequency dictionary, compiled by V. Žilinskienė (1990), was likewise based on only one functional style—journalism; the same Czech methodology was used in its preparation, with extensive collaboration with the Latvians. Over time, despite the fact that Lithuanian cannot boast of this, all kinds of frequency dictionaries appeared around the world: of individual authors’ language, functional styles, regional language varieties, written and spoken language, and so on. Frequency dictionaries of “language in general” became impractical, since specific needs had to be met; consequently, words had to be selected from texts in a particular domain rather than from the entirety of available texts. On the other hand, the problem of compiling frequency dictionaries “entered another dimension” when electronic corpora emerged as American behaviorists and positivists sought examples of usage: electronic technologies were developing rapidly, making it possible not only to collect but also to process huge bodies of text. The beginnings of corpus linguistics are dated to 1961, when Nelson Francis and Henry Kučera announced and began creating the famous electronic corpus of written language, the Brown Corpus (Brown University), while John Sinclair began work on a corpus of the same format, LOB (Lancaster-Oslo/Birmingham)— DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 25 corpus; as early as 1959, Randolph Quirk announced the SEU (Survey of English Usage) corpus of written and spoken language, which was continued by Jan Svartvik, who undertook to create a corpus of spoken language, the LLC (London-Lund Corpus of Spoken English), at Lund University. The size of all the corpora mentioned was then about 1–1.15 million word tokens; more similar corpora appeared later (see Aijmer, Altenberg 1991: 1ff.). The size of second-generation corpora is measured in tens of millions, and that of third-generation corpora in hundreds of millions and billions. The problem of compatibility between computer software and individual work programs remains relevant worldwide: corpora of exceptionally large size—hundreds of millions of words (essentially an electronic form of a library)—have already been compiled, but how can they be used conveniently, how can the most elementary frequencies be counted, and how can one avoid getting lost in an ocean of words? And, ultimately, does quantity always determine quality? The compilation of all kinds of frequency dictionaries has repeatedly shown that half of the words examined occur only once. Therefore, each time one undertakes such work, one needs to consider carefully whether the tasks set really require working with a vast amount of material in search of the rarest words. Sometimes, when certain grammatical or word-formation regularities are being investigated, greatly increasing the size of the sample is indeed irrational; it is entirely appropriate to assess this using statistical criteria as well, since they show the probability of when a sample is sufficient to draw representative conclusions. Particular attention should be paid to which texts make up the sample under study—that is, how the model of the object being investigated is constructed. In Lithuanian linguistics, the issue that now perhaps raises the most questions in this respect is what should be considered standard Lithuanian and what proportions of functional language styles, or of language strata distinguished according to other criteria, should be represented within it. For example, the corpus of Vytautas Magnus University mentioned above was compiled according to one set of sampling principles, whereas the Frequency Dictionary of Contemporary Written Lithuanian was compiled according to another; consequently, the results differ. Statistical data on, for example, individual parts of speech obtained from different corpora must be assessed very carefully and objectively. It has been observed that works of this kind, whether already completed or currently in progress, do not give sufficient attention to the texts examined, yet generalizations are made about “contemporary Lithuanian.” Working with numbers requires precision; otherwise, the conclusions can be misleading. The electronic corpus (1.2 million word tokens) discussed in this article, the Frequency Dictionary of Contemporary Written Lithuanian, belongs to the first-generation corpora (of the Brown and LOB type), with a size of over one million words. It shares another feature with these famous corpora: like few later corpora worldwide, ours is annotated. Computer technology has provided favorable conditions for analytic languages: specialized but fairly widespread programs (the best-known being the University of Oxford’s Oxford Concordance Program, as well as KWIC, KAYE, CLAN, OCP, WordCruncher, etc.) make it relatively easy to compile lists of the most frequent words and provide their concordances (Leech 1991: 10; Utka 2000). Researchers of the Corpus of Contemporary Lithuanian at Vytautas Magnus University generally use programs of this kind, but then only words—that is, the lexicon—can be analyzed, because information about word endings is lacking, and these are, as is well known, especially important in Lithuanian. Thus, things are more complicated for synthetic languages: the programs described above produce lists of word forms rather than headwords. A considerable amount of additional work is required before word forms are normalized and reduced to headwords alone. Such word lists, which also include grammatical analysis, are called annotated; there are few of them worldwide, and they are highly valued because they require substantial labor and are therefore very expensive. 26 | LAIMA GRUMADIENĖ (V. Žilinskienė and I have calculated that in our case, selecting and collecting the texts, entering them using a computer keyboard, carrying out their semi-automated morphological analysis (the so-called LEMMATIZATION), editing them, and loading the data into a computer would require six years of one person’s work). Calculations and subsequent proofreading also take several years, because any inaccuracy in performing mathematical operations can lead to absurdity. In general, especially now that electronic corpora have emerged, frequency dictionaries can be compiled in various ways and in many forms, depending on what one is looking for. The main reason why new frequency dictionaries are continually being compiled, especially those that aspire to be called dictionaries of, for example, contemporary n language, is that the fundamental requirement of linguostatistics, as of statistics in general, is that the sample must be a representative model of the population, or set, under study. In other words, it is continually questioned whether the sample examined really reflects the intended population. For example, a frequency dictionary of contemporary Lithuanian based on the Corpus of Contemporary Lithuanian at Vytautas Magnus University, which consists of as many as 60 million word tokens, would differ from those already published by L. Grumadienė and V. Žilinskienė (1997, 1998), because its results would reflect a population of texts consisting primarily of language from periodicals (70%), non-periodicals (26%), and translated publications (4%) (Marcinkevičienė 2000a: 16). The texts for the aforementioned published dictionaries were selected on an entirely different basis: they comprise original, non-translated texts representing four functional language styles—journalistic, administrative, belles-lettres, and scientific. This approach was adopted because standard Lithuanian, especially written Lithuanian, is still being formed, standardized, and codified in the spirit of the ideas of the Prague Linguistic Circle (incidentally, it should be noted that the term STYLE itself is now defined quite differently by various schools of linguistics). The principles of other frequency-dictionary compilers were followed, primarily those of the Czechs, Russians, and Latvians, who likewise adhered to the conception of standard language—outlined above—as the sum of functional language styles; their experience was applied as well. Thus, as in their work, samples of 300,000 word tokens were selected from four functional styles, since it was considered too early to apply to written Lithuanian the principles used to compile the Vytautas Magnus University corpus (which, incidentally, appeared later than ours), principles closer to the American conception of general (or standard, as is customary there) language. The database of the Frequency Dictionary of Contemporary Written Lithuanian is sometimes criticized for lacking colloquial vocabulary, because people forget that it was formed solely on the basis of public written language: it contains neither private letters and diaries nor inscriptions on walls or the like. Nevertheless, words from spoken language did make it into the dictionary, including not only foreignisms and dialect words but also examples of pejorative vocabulary unacceptable in standard language, for which the dictionary’s authors are continually criticized. The question arises as to what should be considered belles-lettres: these words occurred in literary texts (original texts edited and printed in Lithuania in print runs of more than 100 copies), and the principle of compiling a frequency dictionary is to omit no content word unless it is a proper name, a number, an abbreviation, or a typographical error. Incidentally, the data show that when samples of 300,000 word tokens were selected from each of the four functional styles, 284,687 content words were found in the journalistic style (15,313 so-called “duds”), 279,505 in the administrative style (20,495 excluded from the dictionary for the reasons mentioned), 278,311 in the scientific style (21,689), and 290,561 in the belles-lettres style (9,439). My heart would not let me exclude the proper name Lietuva, so it was the sole proper name admitted to the dictionary as an exception (it turns out to be the 16th most frequent word, the first noun, with a frequency of 5,151 in as many as 614 samples: here the decisive factor was official documents, DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 27 names of institutions and the like, in which this word often occurred in the genitive case). Proper names may also be considered conditionally those of religious and other holidays (often used as common nouns too), e.g., Joninės, Kūčios, Žolinė, or names of the cardinal directions Rytai, Vakarai, Pietūs, Šiaurė. In keeping with the principle of representativeness, only texts belonging to the pure form of each of the aforementioned styles were selected: borderline textbook texts, scientific journalism, journalism aimed at a narrow readership, drama, poetry, and the like were excluded. The publicistic texts were selected by Vida Žilinskienė; they comprised texts from 19 edited national daily and weekly newspapers and widely read magazines from 1994–1995, representing five subject groups: 1) politics, 2) industry and agriculture, 3) education and law, 4) the arts, and 5) entertainment (sports, horoscopes, weather forecasts). The texts in the administrative style were selected by Pranas Kniūkšta and Danutė Vainauskienė from publications and documents dated 1990–1995, grouped under seven subjects: 1) laws and accompanying documents and commentaries, 2) legal documents, 3) economics and business reference books and documents, 4) political science publications and party documents, 5) standards and statistical publications, 6) information sources (bibliographies, catalogues, dictionaries), and 7) documents from particular social fields (national defence, social security, education, culture, etc.). The scientific-style texts were selected by Irena Andriukaitienė from scientific publications issued in Lithuania in 1990–1995 or works approved by scientific institutions, across eight subject groups (fields of science): 1) agricultural sciences, 2) natural sciences, 3) humanities, 4) mathematics, 5) medicine, 6) social sciences, 7) technology, and 8) theology. The literary texts were selected by Laima Grumadienė from prose works of fiction published in Lithuania in 1985–1990 (with a few exceptions); their six “thematic” groups were classified according to the writers’ birthplaces: 1) Samogitians, 2) eastern Highlanders, 3) western Highlanders, 4) southern Highlanders, 5) townspeople, and 6) emigrants (the latter’s actual birthplaces were not taken into account). It was decided that these texts could represent contemporary written Lithuanian. From them, a 1,000-word-occurrence segment, called a sample, was selected at random (pages and lines were chosen using a table of random numbers) and entered using a computer keyboard (about two hours’ work). The samples were then lemmatized, i.e., morphologically analyzed using MAN (Morfologinio Analizavimo ir Normalizavimo), a program developed by Vytautas Zinkevičius. The author adapted it specifically for this project from his own earlier computer program for correcting the spelling of words; in 2000, at the request of Vytautas Magnus University, he improved it somewhat, upgraded it to a higher version, and named it the lemmatizer (Zinkevičius 2000). The process looked roughly like this: a computer printout was produced, and the unnecessary lines (underlined by me) had to be “thrown out.” For example, the printout of a text segment already processed by a person, not just a machine (…) manęs visa ši komedija (…) looked like this: FIG. 1. EXAMPLE OF TEXT LEMMATIZATION I think me pron. personal, singular, accusative; me verb, past participle, active voice, negative, masculine, singular, nominative 28 | LAIMA GRUMADIENĖ entire entire pron. feminine, singular, nominative entire pron. feminine, singular, instrumental entire pron. feminine, singular, genitive entire pron. adverb this this pron. indefinite, feminine, singular, nominative comedy comedy noun, feminine, singular, nominative comedy noun, feminine, singular, instrumental comedy noun feminine singular nominative The first line retains the restored headword; the second (the remaining, uncrossed-out line) gives the word form with its grammatical characteristics. Processing 1,000 words in this way takes about 4 hours, but there is probably still no faster way to analyze Lithuanian words morphologically. After that, homonyms and homophones are sought, the samples are corrected using Vytautas Zinkevičius’s MAN-K program, and they are merged into the computer—all this takes another hour. In total, 1,200 samples of 1,000 word occurrences each were processed in this way. After this enormous amount of work had been completed, the data were further processed using calculation programs prepared by Valentinas Černiauskas and Snieguolė Meškauskienė (both from the Institute of Mathematics and Informatics): various frequencies were calculated. Three dictionaries were prepared for publication; two have already been published (Grumadienė, Žilinskienė 1997; 1998), and the third will, we hope, see the light of day this year. The first two dictionaries are frequency lists of headwords. The first is arranged in descending order of frequency, the second alphabetically. They contain 38,337 different words. The third, a dictionary of word forms, is several times larger (31 printed sheets), contains approximately 150,000 word forms, and reflects the use of different words and of the same words in various grammatical forms (it should be presented both alphabetically and in reverse order). Several dozen more frequency dictionaries, arranged in various ways, have been prepared for electronic use: for all four styles, for the thematic groups within each style, and so on (both headwords and word forms). A comparison of some statistical data already shows that the greatest variety of distinct nouns is used in written Lithuanian: nouns account for as much as 45.25% of all distinct words; verbs and verb forms are plentiful, at 32.65%; adjectives make up a significant share, at 15.31%; and 5.03% of distinct words are adverbs. Other parts of speech each account for less than half a percent (0.88% altogether). This does not, however, mean that their distribution in texts is the same. Another measure—the frequency of the same words in texts—shows that the percentages of the main parts of speech and adverbs decline considerably in favor of function words and pronouns, which together make up more than a quarter of words in texts, at 26.71%. Attention should be paid not only to frequency, but also to the number of samples in which a word or word form occurs: this makes it possible to determine whether a word’s frequency was caused by a single text that happened to contain many repetitions of it, or whether its frequency is not accidental after all, since its dispersion across different samples is uniform. DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 29 Lexical diversity is assessed using the iteration coefficient (the ratio of word occurrences to headwords), which in the dictionary mentioned is 20.06 (the smaller this value, the greater the number of distinct words found). The compiled but unpublished dictionaries of headwords and word forms for each of the four functional styles (V. Žilinskienė has already written about some of the linguistic-statistical features observed in them (1998, 2001, in press, etc.)) show that, for example, the vocabulary of fiction is twice as diverse as that of texts in the expository style (the iteration coefficients are 9.53 and 20.78, respectively). Most interesting here is that the iteration coefficient for all four styles taken together differs relatively little from that of the expository style, but greatly from that of fiction (the coefficient is 10.51 for journalism and 12.63 for science). These indicators are a compelling argument when deciding, for example, on the principles for compiling a dictionary of standard language: what proportions should be assigned to the functional styles; whether the practice of associating standard language with functional styles should be continued at all; and, if so, the statistical indicators of functional styles must be taken into account, among other things. Interesting data were obtained by comparing the list of the most frequent words compiled on the basis of the Vytautas Magnus University corpus (in it, words were recorded and counted in the forms in which they occur in texts, without restoring their headword forms, unlike in the Frequency Dictionary of Contemporary Written Lithuanian). A comparison of the 50 most frequent uninflected words and inflected word forms in a 60-million-word Lithuanian corpus (Utka 2000: 278) with the first 50 headwords in the Frequency Dictionary (Grumadienė, Žilinskienė 1997: 3) makes it clear that they are essentially the same. The first frequency list contains slightly fewer distinct words, because it includes various forms of the same word, e.g., was, to be; he, his, their, her, they, whereas in the second list, that is, in the dictionary, these are represented by to be and he, respectively. In accordance with the principles for compiling this type of dictionary, abbreviations were not included in the Frequency Dictionary, but comparison of the abbreviations m, d, a found in the aforementioned list of the most frequent words—that is, year, day, age—with the data in the Frequency Dictionary shows that the first two words are also frequent in the dictionary, ranking 30th and 57th, respectively (words with the same frequency are assigned the same rank), whereas age appears only in the third hundred. Incidentally, both the abbreviation and the full word are used quite often, and with somewhat different meanings. As confirmed by the practice of compiling frequency dictionaries for many synthetic languages, the most frequent words in the Frequency Dictionary of Contemporary Written Lithuanian are some conjunctions (and, that, or, how, whether, but, and, when, etc.), prepositions (into, from, with, from, about, after, behind, because of, through, near, etc.), pronouns (he, that, I, which, this, all, other, own, such, what, you, etc.), particles (and, not, only, etc.); the most popular verbs are to be, to be able, to have, not to be, to need, to know, to determine, etc., and the nouns Lithuania, work, person, year, etc. The most frequent numerals are one, first, two, while the adjectives included among the first hundred most frequent words are big, new, high, various, important, main. Incidentally, data from analytic languages also show that, for example, in English texts the pronoun I, the conjunction and, and the article the account for 10% of texts (Hillerich www: eduplace.com). The question of international words is also important: how many of them, and which ones, should be included in a dictionary of standard language? It may be useful to bear in mind that, for example, 27.52% of the headwords in the Frequency Dictionary of Contemporary Written Lithuanian (1998, XXXIV–XXXVI) were found to be international words, and that they accounted for 13.34% of the text; the lists of those words themselves may also prove useful. 30 | LAIMA GRUMADIENĖ In preparing to write a new grammar of the Lithuanian language, it is important to know the statistics of word forms as well: which types of words inflect in what ways and why, and why the range of grammatical forms is large for some words but not for others. Such data are abundant in the dictionary of word forms currently being prepared for publication. V. Žilinskienė (2001) has already published interesting data concerning the journalistic style alone. It turns out that three noun cases account for 82.29% of all noun occurrences (gen. – 38.91%, nom. – 26.62%, acc. – 16.76%; and the fact that, for example, the dat. is used in only 3.88% of cases may already indicate a tendency toward analyticity?). Adjectives, incidentally, are most often used in the nom. case – in 36.93% of cases. It has been established that singular nouns predominate – 68.86% of all occurrences. Relatively few gerunds (2.34%) and half-participles (1.84%) were found in journalistic texts, whereas participles accounted for as much as 24.46%; the infinitive was used in 15.85% of cases. Interesting and unusual results, especially for teachers and specialists preparing textbooks, curricula, and dictionaries, were obtained by examining the three Minimal Lithuanian–Russian Dictionaries for State Language Proficiency Qualification Categories prepared and published by V. Žilinskienė and L. Grumadienė (1998; 1999a; 1999b), as well as six other analogous Lithuanian–English and Lithuanian–Polish dictionaries that have been prepared but not yet published. Here, it is worth noting once again that the samples were drawn from written Lithuanian texts, so spoken-language words may be missing. These dictionaries are intended for learning written language, since the qualification requirements for state-language proficiency concern the use of the state language in public, primarily in official situations, especially in state institutions. Consequently, non-colloquial but neutral and specialized vocabulary is needed here: users are required to be able to read and write business documents and understand media texts. In the future, minimal dictionaries of spoken language should also be prepared. In Lithuania, Vytautas Šernas’s (1994) textbook How to Learn Languages More Rationally has been published; it also contains a minimal bilingual English–Lithuanian dictionary: a translation of Ch. K. Ogden’s famous 1942 publication Basic English, comprising 850 words (it can also be found in volume 3 of Encyclopaedia Britannica). As time passes, some words become obsolete and others emerge, so the list of those 850 words may change accordingly. In setting the requirements for the three qualification categories of proficiency in the state Lithuanian language, the opinion of the authors of the Frequency Dictionary of Contemporary Written Lithuanian regarding the scope of minimal dictionaries was also taken into account: it was decided that knowledge of about 800 words would meet the requirements of Category I, about 1,200 words those of Category II, and about 2,500 words those of Category III. The decision was based on Ogden’s widely accepted postulate concerning 850 words, on frequency studies of Lithuanian and other languages, and on analogous practices in other countries. That these calculations are not arbitrary is also evident from the three tables in the appendix, which clearly show what proportion of texts is covered by the 800, 1,200, and 2,500 most frequent words in written Lithuanian, respectively: 56.66%, 65.70%, and 78.04%. It is now becoming customary for all major dictionaries to provide information about the most frequent words in a language. For example, in the English COBUILD dictionary (www.antimoon.com), the 680 most frequent words are marked with one to four diamonds, respectively. In Longman’s English dictionary, the 3,000 most frequent words are marked S1, S2 (S – spoken English), W1, W2 (W – written English), respectively. In Macmillan’s English dictionary, the most frequent words are printed in red, while the frequency of other words is indicated by one to three stars, respectively. The time is approaching when Lithuanian lexicography will apparently have to take up this practice too. DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 31 4. CONCLUSIONS There are few electronic databases of contemporary Lithuanian, the largest being V. Žilinskienė and L. Grumadienė’s Frequency Dictionary of Contemporary Written Lithuanian (morphologically annotated and statistically processed, with 1,200,000 word occurrences) and Vytautas Magnus University’s Corpus of the Lithuanian Language (unannotated, with approximately 60,000,000 word occurrences). Lithuanian studies still lack appropriate practices for using both electronic databases of contemporary language and linguostatistical publications. Above all, students in Lithuanian higher education institutions are not sufficiently encouraged to work in this way; but there is another side to the issue: too few electronic databases of Lithuanian are publicly accessible. One reason for this lag is that electronic databases are not recognized as authored scholarly works. One of the most pressing problems in Lithuania concerning electronic databases is the weakness of the principles guiding their compilation: electronic databases are often intended to be used merely as texts presented in electronic form, without providing for the application of computer technologies to process the data. 32 | LAIMA GRUMADIENĖ TABLE 1: FREQUENCY OF WORDS BY LINGUISTIC CATEGORY AND OVERALL IN THE MINIMAL DICTIONARY FOR STATE LITHUANIAN LANGUAGE PROFICIENCY Rank Frequency noun verb adjective adverb other 1–10 39582-7001 – 21262 – – 52221 14,38% 35,31% 11–20 6877-4274 – 10582 – – 21221 19,96% 40,02% 21–30 4113-2989 10763 – – – 11048 28,95% 29,71% 31–40 2950-2637 – – – 5619 – 20,35% 41–50 2631-2053 – 6843 2053 4782 2113 30,50% 9,15% 21,32% 9,42% 51–60 2015-1781 9284 – – 3867 – 45,22% 18,83% 61–70 1754-1557 10014 1739 – – 1593 60,41% 10,49% 9,62% 71–80 1548-1461 6036 3046 1507 3009 – 40,04% 20,20% 10,00% 19,96% 81–90 1460-1373 1422 2828 1373 2888 2792 10,03% 19,94% 9,68% 20,36% 19,68% 91–100 1351-1148 7421 1305 – 1232 1159 54,63% 9,61% 9,07% 8,53% 101–150 1442-905 23551 12523 7419 3116 943 43,03% 22,88% 13,55% 5,69% 1,72% 151–200 900-706 22744 10742 870 4203 1583 52,48% 24,79% 2,01% 9,70% 3,65% 201–250 702-564 14059 6315 4906 3178 3148 42,09% 18,90% 14,69% 9,51% 9,42% 251–300 562-462 17325 7962 1451 3605 1041 49,59% 22,79% 4,15% 10,32% 2,98% DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 33 LINGUISTIC ANALYSIS OF THE MINIMAL DICTIONARY FOR QUALIFICATION CATEGORIES: other verb adjective adverb noun total – 25050 49371 – – 147904 16,93% 33,38% 12,33% – – 16902 4320 – 53025 31,87% 8,15% 4,42% – 7156 4102 4113 – 37182 19,25% 11,03% 11,06% 3,10% 2641 8282 8281 2791 – 27614 9,56% 29,99% 29,99% 10,11% 2,30% 2116 4526 – – – 22433 9,43% 20,18% 1,87% – 3589 3793 – – 20533 17,48% 18,47% 1,71% 1557 1673 – – – 16576 9,39% 10,09% 1,38% – – 1478 – – 15076 9,80% 1,26% – 1427 – 1454 – 14184 10,06% 10,25% 1,18% – – 1237 1230 – 13584 9,11% 9,05% 1,13% 1007 2164 2986 1028 – 54737 1,84% 3,95% 5,46% 1,88% 4,56% 738 – 900 1557 – 43337 1,70% 2,08% 3,59% 3,61% – – 574 1225 – 33405 1,72% 3,67% 2,78% – 1035 512 2007 – 34938 2,96% 1,47% 5,74% 2,91% 34 | LAIMA GRUMADIENĖ Rank Frequency nouns verbs adjectives adverbs pronouns 301–350 460-374 19646 10424 2420 2073 393 54,28% 28,80% 6,69% 5,73% 1,09% 351–400 373-313 18504 7550 3381 1347 654 53,04% 21,64% 9,69% 3,86% 1,87% 401–450 312-260 19156 7865 3976 522 304 57,20% 23,49% 11,87% 1,56% 0,91% 451–500 259-205 21718 9987 4806 2486 471 53,17% 24,45% 11,76% 6,08% 1,15% The minimum vocabulary for the State Lithuanian Language I qualification category (about 800 TABLE 2: WORD FREQUENCY BY PART OF SPEECH AND OVERALL FOR THE STATE LITHUANIAN LANGUAGE PROFICIENCY II QUALIFICATION CATEGORY Rank Frequency nouns verbs adjectives adverbs pronouns 501–550 204-155 22494 13153 7023 2481 525 46,97% 27,46% 14,66% 5,18% 1,10% 551–600 154-106 30834 17197 6221 4419 151 50,91% 28,39% 10,27% 7,29% 0,25% The minimum vocabulary for the State Lithuanian Language II qualification category (about 1200 TABLE 3: WORD FREQUENCY BY PART OF SPEECH AND OVERALL FOR THE STATE LITHUANIAN LANGUAGE PROFICIENCY III QUALIFICATION CATEGORY Rank Frequency nouns verbs adjectives adverbs pronouns 601–650 105-56 45751 20328 12145 5822 433 52,31% 23,25% 13,88% 6,66% 0,50% The minimum vocabulary for the State Lithuanian Language III qualification category (about 2500 DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 35 abbr. pres. conj. fut. inf.; imp. Total 426 – – 812 – 36194 1.17% 2.24% 3.02% – – 1407 2043 – 34886 4.03% 5.87% 2.91% 276 – 272 1118 – 33489 0.82% 0.81% 3.34% 2.79% 255 224 210 700 – 40857 0.62% 0.55% 0.51% 1.71% – 3.40% words) covers 56.66% of the text (sample of 1.2 million word occurrences). 56.66% LINGUISTATISTICAL ANALYSIS OF THE MINIMAL DICTIONARY CATEGORIES: abbr. pres. conj. fut. inf.; imp. Total 530 342 177 795 371 47891 1.11% 0.71% 0.37% 1.66% 0.78% 3.99% 520 275 132 713 109 60571 0.86% 0.45% 0.22% 1.18% 0.18% 5.05% words) covers 65.70% of the text (sample of 1.2 million word occurrences). 9.04% Total: (=65.70%) LINGUISTATISTICAL ANALYSIS OF THE MINIMAL DICTIONARY CATEGORIES: abbr. pres. conj. fut. inf.; imp. Total 460 341 372 1697 102 87446 0.52% 0.40% 0.42% 1.94% 0.12% 7.29% words) covers 78.04% of the text (sample of 1.2 million word occurrences). 7.29% Total: (=78.04%) 36 | LAIMA GRUMADIENĖ LITERATŪRA AJMER, K., ALTENBERG, B. 1991: Introduction. Ajmer, K., Altenberg, B., eds., English Corpus Linguistics. Studies in Honour of Jan Svartvik, London–New York: Longman. ALAMEDA, J. R., CUETOS, F. 1995: Diccionario de frecuencias de las unidades lingüísticas del castellano, Universidad de Oviedo. BITINIENĖ, A. 1983: Mokslinis stilius, Vilnius: Aukštojo ir specialiojo mokslo ministerija. BITINIENĖ, A. 1997: Funkciniai stiliai: sakinio ilgis ir struktūra, Vilnius: Vilniaus pedagoginio universiteto leidykla. BITINIENĖ, A. 1998: Administracinis stilius ir jo sakinio ilgis. Kalbotyra 47 (1), 17–28. EATON, H. 1934: Comparative frequency list on the first thousand words in English, French, German and Spanish. Coleman, A., ed., Experiments and Studies in Modern Language Teaching, Chicago. GARCIA HOZ, V. 1953: Vocabulario usual, comun y fundamental. Madrid: Consejo Superior de Investigaciones Cientificas – Instituto „San José de Calasanz“. GRUMADIENĖ, L., ŽILINSKIENĖ, V. 1997: Dabartinės rašomosios lietuvių kalbos dažninis žodynas (mažėjančio dažnio tvarka), Vilnius: Matematikos ir informatikos institutas, Lietuvių kalbos institutas. GRUMADIENĖ, L., ŽILINSKIENĖ, V. 1998: Dabartinės rašomosios lietuvių kalbos dažninis žodynas (abėcėlės tvarka), Vilnius: Matematikos ir informatikos institutas, Lietuvių kalbos institutas. HILLERICH, R., www: eduplace. com: High-frequency words and vocabulary. JAKUBAITĖ, T. e.a. 1966–1976: Latviešu valodas biežuma vārdnīca, 1.–4. sējumi, Rīga: Zinātne. JOSSELSON, H. 1953: The Russian Word Count and Frequency Analysis of Grammatical Categories of Standard Literary Russian, Detroit: Periodical Service Co. JUILLAND, A., RODRIGUES, E. Ch. 1964: Frequency Dictionary of Spanish words, The Hague: Mouton. JUILLAND, A., EDWARDS, P. M., JUILLAND, I. 1965: Frequency Dictionary of Roumanian Words, The Hague: Mouton. JUILLAND, A., TRAVERSA, V. 1973: Frequency Dictionary of Italian Words, The Hague: Mouton. KAEDING, F. 1898: Häufigkeitswörterbuch der deutschen Sprache, Steigliz bei Berlin (Selbstverlag). KEINYS, S., red. 1993: Dabartinės lietuvių kalbos žodynas, Vilnius: Mokslo ir enciklopedijų leidykla. KEINYS, S., red. 2000: Dabartinės lietuvių kalbos žodynas, Vilnius: Mokslo ir enciklopedijų leidykla. KENNEDY, G. 1998: An Introduction to Corpus Linguistics, London–New York: Longman. KORSAKAS, J. 1991: Lietuvių kalbos inversinis žodynas, Kaunas: Šviesa. KRUOPAS, J., red. 1954: Dabartinės lietuvių kalbos žodynas, Vilnius: Valstybinė politinės ir mokslinės literatūros leidykla. KRUOPAS, J., red., 1972: Dabartinės lietuvių kalbos žodynas, Vilnius: Mintis. LEECH, G. 1991: The state of the art in corpus linguistics, Ajmer, K., Altenberg, B., eds., English Corpus Linguistics. Studies in Honour of Jan Svartvik, London–New York: Longman. LÖNGREN, L. 1993: Частотный словарь современного русского языка. Uppsala. (Acta Universitatis Upsaliensis, Studia Slavica Upsaliensia 32.) MARCINKEVIČIENĖ, R. 2000a: Tekstynų lingvistika (teorija ir praktika). Darbai ir dienos 24, 7–64. MARCINKEVIČIENĖ, R. 2000b: Patterns of word usage in corpus linguistics. Kalbotyra 49 (3), 71–80. MAŽEJKA, N. S. (Мажэйка, Н. С.), SUPRUN, A. Ja. 1976: Частотны слоўнік беларускай мовы (мастацкая проза), Мінск: Выдавецтва БДУ імя У. Леніна. MIKELIONIENĖ, J. 2000: Naujoji lietuvių kalbos leksika (1991–1996 m. kompiuterinio periodikos tekstyno pagrindu). Daktaro disertacija, Kaunas: Vytauto Didžiojo universitetas. MURMULAITYTĖ, D. 1998: Žodžių sąsajos žymėjimas „Dabartinės lietuvių kalbos žodyne“ (santrumpos žr. vartojimas). Lietuvių kalbotyros klausimai 39, 240–247. MURMULAITYTĖ, D. 1999: „Bendrinės lietuvių kalbos žodyno“ kompiuterizavimo klausimu. Lituanistica 2 (38), 92–98. RYKLIENĖ, A. 2000: Bendravimas internetu: kalbėjimas rašant. Darbai ir dienos 24, 99–107. DABARTINĖS KALBOS DAŽNINIS ŽODYNAS IR JO BAZĖ | 37 ROBINSON, D. F. 1976: Lithuanian Reverse Dictionary, Michigan: Slavica Publishers Inc. ROZENCVEJG, V. Ju. 1986: Опыт создания национальных лексикографических служб за рубежом. Караулов Ю. Н., ред., Машинный фонд русского языка: идеи и суждения, Москва: Наука, 75–83. SAVICKIENĖ, I. 1999: Lietuvio vaiko daiktavardžio morfologija. Daktaro disertacija, Kaunas: Vytauto Didžiojo universitetas. ŠERNAS, V. 1994: Kaip racionaliau mokytis kalbų (Basic English), Vilnius: Žodynas. ŠTEJNFEL’DT, E. A. 1963: Частотный словарь современного русского литературного языка, Таллин. TĚŠITELOVÁ, M. e.a. 1983: Frekvenční slovník češtiny věcného stylu, Praha: Academia–Nakladatelství Československé Akademie Věd. THORNDIKE, E., LORGE, L. 1944: A Teacher’s Word Book of the Twenty Thousand Words Found Most Frequently and Widely in General Reading for Children and Young People, New York: Appleton-Century Crofts. TODOROVA, E., PANČOVSKA, R. 2001: Честотен речник на българската публицистика (1944–1989), София: Пенсофт. UTKA, A. 2000: Kalbinė įranga ir jos galimybės. Darbai ir dienos 24, 275–285. VACAR, N. P. 1966: A Word Count of Spoken Russian. The Soviet Usage, Ohio: Ohio State University Press. VANDER BEKE, G. 1929: French Word Book, New York: Macmillan. ZASORINA, L. N., ред., 1977: Частотный словарь русского языка, Москва: Издательство «Русский язык». ZINKEVIČIUS, V. 2000: Lemiuoklis – morfologinei analizei. Darbai ir dienos 24, 246–273. ŽILINSKIENĖ, V. 1990: Lietuvių kalbos dažninis žodynas, Vilnius: Mokslas. ŽILINSKIENĖ, V. 1995: Atgalinis dabartinės lietuvių kalbos žodynas, Vilnius: Matematikos ir informatikos institutas. ŽILINSKIENĖ, V. 1998: Dažninis dabartinės rašomosios lietuvių kalbos žodynas (pirmieji tyrimo rezultatai). Lietuvių kalbotyros klausimai 39, 219–227. ŽILINSKIENĖ, V., GRUMADIENĖ, L. 1998: Valstybinės kalbos mokėjimo kvalifikacinių kategorijų lietuvių–rusų kalbų minimalusis žodynas (pirmoji kvalifikacinė kategorija), Kaunas: Šviesa. ŽILINSKIENĖ, V., GRUMADIENĖ, L. 1999a: Valstybinės kalbos mokėjimo kvalifikacinių kategorijų lietuvių–rusų kalbų minimalusis žodynas (antroji kvalifikacinė kategorija), Vilnius: UAB „Nacionalinių tyrimų centro“ leidykla. ŽILINSKIENĖ, V., GRUMADIENĖ, L. 1999b: Valstybinės kalbos mokėjimo kvalifikacinių kategorijų lietuvių–rusų kalbų minimalusis žodynas (trečioji kvalifikacinė kategorija), Vilnius: UAB „Nacionalinių tyrimų centro“ leidykla. ŽILINSKIENĖ, V. 2001: Lietuvių ir latvių kalbų publicistikos leksikos ir morfologijos raida statistiniu požiūriu. Linguistica Lettica 9, 155–168. ŽILINSKIENĖ, V. spausd.: Lietuvių kalbos dalykinio stiliaus morfologijos statistinės charakteristikos ir jų lyginimas su atitinkamomis publicistikos charakteristikomis. Lituanistica. Laima Grumadienė Lietuvių kalbos institutas P. Vileišio g. 5, LT–2055 Vilnius, Lietuva [email protected] Gauta 2002 04 19