Technologies of deep text analysis in musicology

Michkov Pavel Aleksandrovich ORCID: 0000-0001-8227-7190 PhD in Art History Associate Professor, Department of Music Theory, Federal State Budgetary Educational Institution of Higher Education "Novosibirsk State Conservatory named after M.I. Glinka" 630099, Russia, Novosibirskaya oblast', g. Novosibirsk, ul. Sovetskaya, 31, of. 224 p.michkov@nsglinka.ru

At the moment, information support and description of scientific papers are actively changing. Information search is expanding and improving. Increasingly, the standards for describing the corpus of scientific literature suggest the use of its automatic processing. These conditions are also applied to works in the field of musicology. So, in many scientific journals about the musical art and on the websites of publishing houses, the rules for the design of the work for publication are set out: the rules include a number of parameters: the author's surname, the title of the work, an abstract (usually in two languages) and a set of keywords. Note that the set of keywords is compiled by the author of the study. It is the author who can most accurately determine those lexical units in which the meaning of the text is concentrated. On the other hand, this approach somewhat veils the phenomenon of the ambiguity of the text, the invariance of its semantic field. The author consciously encloses the text in a certain framework, set by the parameters of the key values in the form of a set of words and phrases. Based on this, the subject of this article is the process of forming a set of key phrases by means of computer assisted technologies.

Thus, when describing a document in automated library systems, key values are also introduced. If there is no author's set of such meanings, then those words and phrases that are used as the title of the source or the titles of the sections of the work are crucial. This means that during the search process, documents in which the keywords specified by the user are nearby will be prioritized, that is, they will be ranked above all other documents. Thus, it turns out that this method, which is currently the simplest and most convenient, ultimately does not guarantee the adequacy of the transmission of the content of the text with keywords.

More promising is the process of user involvement in the search operation – in the stages of setting up the information search scheme. In this case, the process of searching for the necessary set of keywords turns out to be personalized. So, the dialog structure "researcher – a computer program for text analysis" can be used:

a) in the process of preparing the study for publication;

b) in the work of the bibliographer when cataloguing any publication;

c) in the process of analyzing the structure of a scientific problem or the hidden properties of text arrays.

A number of authors investigate the methodology of engagement processes. So, with regard to the first two points ("a", "b"), T. Volkova suggests applying a similar procedure, arguing that "for digitized paper documents, it will be necessary to conduct a semantic analysis of texts and on the basis of such analysis to obtain a compressed text that is keywords and phrases" ^{[1, p. 285]}. However, her research does not talk about how such an analysis of the text will be carried out. In both cases, the use of deep text analysis programs may be one of the solutions to the problem. Let's consider the capabilities of the TextAnalyst program using the example of a comparative analysis of a set of keywords proposed by the authors themselves with the results of the analysis of the same texts performed by the TextAnalyst program.

In the 70s of the XX century, in the works of A. Mol ^[2], I. D. Rud, I. I. Zuckerman ^[3], the possibility of using an information approach in musicology began to be considered in connection with the development of information systems. Musicology has extrapolated the basic laws of information theory into music science, finding similarities with them in the doctrine of musical formation and the sociology of musical culture. It is obvious that the development of the information approach in musicology is directly related to text structures. Performing various operations with the text formed new methods of research of scientific information, organically including them in musicology. This is confirmed by the appearance of works on the study of musicological research creativity. A striking example is the work of T. I. Naumenko ^[4] on the analysis of the style of verbal texts of musicology.

Articles from the journals "Musicology" and "Music and Time" were chosen as the material for the experiment. This choice was due to the subject matter of the publications, which covers a wide range of musicological problems (questions of theory, history of music, musical ethnography, paleography, pedagogical issues and problems of musical performance are covered), as well as the fact that since 2009, authors must accompany the text of the article with a set of keywords. It should be noted that in these periodicals, the authors of the publications, together with the text of the work, also attach a brief annotation.

It is assumed that the set of keywords given by the authors of the works may differ or coincide with the results of computer analysis of the text. Stylistic or thematic kinship of studies located in the same group is also possible. In addition, we take into account that in the process of subsequent search for a scientific article by the reader, "viewing annotations and keywords facilitates the task of getting acquainted with new products, but, of course, does not solve the problem of collecting and in-depth study of literature on a particular topic" ^{[5, p. 99]}.

As the first example illustrating the results of an experiment conducted to determine a group of keywords by means of a deep text analysis program, we present articles from the magazine "Music and Time". The examples were taken in pairs due to their thematic and some stylistic heterogeneity. Both articles are devoted to the study of problems in the field of music science: the work of I. Yu. Proskurina's "The Last Work of A.K. Glazunov: a Musicological Investigation" ^[6] is in the field of historical musicology and is devoted to the study of a musical work of the late period of A. K. Glazunov's creativity. The scope of the problems identified in T. P. Petrova's publication "The Ceremonials" by D. Ligeti: Between Meaning and Absurdity" ^[7] is at the intersection of the fields of historical and theoretical musicology.

In the diagrams that represent the results of the analysis of articles (Example 1), it is noted that the set of keywords proposed by the authors of the analyzed articles intersects with the results of the text analysis performed using the TextAnalyst program. Meanwhile, in most cases, the range of key values generated by the program is much wider than the set presented by the researchers:

Example 1

Comparative scheme of keyword set analysis

articles by G. P. Petrova "Ceremonials" by D. Ligeti: between Meaning and Absurdity"

With regard to point "c", let's consider another type of text analysis – sets of words and phrases formed on the basis of the specific weight of text elements. Perhaps, in this case, new perspectives of text analytics arise.

In the periodical "Musicology", groups of materials are formed according to the structural composition of the publication – according to the headings of the journal. Such an organization makes it possible to combine thematically related studies into joint sections. The current system of headings has been established in the periodical under study. A separate section includes the heading "Works of young scientists", which offers research by graduate students, often in different thematic areas, but, meanwhile, combined into one group. In addition to these headings, the journal presents the "Conferences" section, which highlights the events of past scientific meetings and discussions.

The thematic content of the issues of the magazine for the calendar year covers the work of a wide range of composers of different stylistic eras and trends. Journalistic materials cover facts related to the musical activities and biographies of L. Beethoven, M. P. Mussorgsky, B. Bartok, G. V. Sviridov, F. Karaev, V. A. Gavrilin, H. Distler and others. Works on ethnomusicology explore the traditions of creativity of Tajiks, Tuvans, Indians, Buryats, Vietnamese and Slavic peoples. The research topics of young scientists are diverse – from the stylistic features of I. F. 's creativity . Stravinsky, E. I. Podhajtsa, A. Dvorak, A. Piatsolla to the characteristics of genres of children's musical folklore and the problems of Russian musical constructivism.

After the necessary preparation and processing of the text, its analysis is performed using a computer program ^[1]. In the process of analysis, the text is divided into segments, between which relationships are established. Based on the weight of the relationships in the text, the program determines a set of the most significant words and phrases. Such a set can be conditionally considered a set of keywords and expressions of the source text. In addition, a scheme of the frequency distribution of words in the text is proposed, based on the number of their repetitions. Thus, formally, the purpose of such an analysis is to build a tree of dependencies between words and phrases in the text. Proposals are presented in the form of branches with a single root vertex. Since one word form can correspond to several grammatical forms of a word, it is necessary to verify the results of segmentation during the analysis.

The method in which the author's ideas about the original scientific problem are built up in the form of a structure leads to the identification of the context of the work and the alignment of potential areas-sources of information. According to the correct remark of N.S. Bazhanov: "in order to improve the search for information, it is necessary to more fully and holistically present the structure of information search" ^{[8, p. 3]}. In terms of studying the material by means of deep text analysis programs, there is a tendency to identify potential contextual areas at the level of primary analysis. There is a possibility of finding hidden properties of text arrays.

In T. B. Budaeva's article "The Vocal side of the Beijing Opera" ^[9], the abstract was excluded from the general body of the analyzed text. The results of the analysis showed that the words "makeup" and "paint" are often interacting expressions at a fairly close distance (Example 2):

 Example 2.

View of the abstract of T. Budaeva's article "The vocal side of the Beijing Opera"

There are none in the group of keywords in the author's annotation, but after reading the article it turns out that the spectacular and coloristic performances of the Beijing Opera are represented by bright makeup and colors, while less attention is paid to the musical side in the research literature. The conclusion suggests itself that in the case of searching for sources whose content touches on the topic of theatricality of Chinese theater productions, the publication in question will be excluded from the search results, since the meaningfully important meanings ("makeup", "paint") are absent from the keywords.

It is important to note that in this case, when using a deep text analysis program in their activities, the author of the publication has the opportunity to either supplement the existing list of keywords, or reformulate it. In any case, the researcher is invited to take another look at his text from a slightly different angle, given the opportunity to identify new sides and identify new meanings that he probably had not noticed before.

As a rule, software products of this kind determine the quantitative characteristics of the most frequently used words and phrases. In the presented analysis of musicological articles, in addition to the frequency distribution of phrases, a scheme is presented, constructed in the order of the semantic meaning of phrases in the text. This structure is determined by the number of repetitions throughout the material. The role of repetition in the humanities has recently become the subject of close attention of researchers. The principles of repetition are studied in detail in the works of literary critics – Y. Lotman ^[10] and S. A. Reiser ^[11].

Turning to the analysis of repetitive elements in the structure of musicological texts allows us to make several assumptions about the reasons for their appearance and role in the process of reader perception. If two concepts are in the same sentence, then there is an interaction between them, which is explicitly or covertly narrated by this text. The increased frequency of the use of a pair or group of terms in a paragraph also indicates their relationship, which needs to be disclosed, explained, discussed.

Thus, one of the important factors influencing the quantitative composition of repeated phrases may be the stylistic attitude of the author to the specific text he creates. It is known that the style of a scientific work can differ significantly from a journalistic, official-business or artistic one. The stylistic technique of repetition is one of the main means of accentuation in the text, its compositional organization. Consequently, elements characteristic of one style may not be involved in another. Nevertheless, there are examples of a subtle fusion of different stylistic intentions within a single text.

Another factor affecting the repetition of words and phrases is the psychological aspect. The author can quite deliberately use the repetition of certain words and expressions to emphasize the focus of the reader's attention to specific aspects and elements of the covered topic. In the text, such techniques perform a double function: on the one hand, they act as a method of compositional design of the material, a means of developing the main thematic elements, on the other hand, they contribute to achieving communicative goals, focusing attention, forming a certain evaluative attitude to the text.

In addition to the obvious techniques – stylistics and intentional concentration of text structures, the "body" of the text may contain key elements that make up a separate thematic area. And finally, the most important thing. The repetition of the use of words and phrases is due to semantic, semantic relationships that objectively exist between them in the text. The identification of such interactions with the help of assisting computer programs for text analysis opens up new horizons, new opportunities in the activities of a musicologist.

Thus, it can be noted that the information approach offers the musicologist another option for the study of the phenomenon of musical art he is studying. Such a problematic "slice" turns out to be saturated with other information content, which makes it possible to find hidden patterns of development of certain processes in the future. Of course, like any other, this method has its own limits of use and is not a universal technology at all. It is impossible to outline the zone of its action with an exact line marking the boundaries of truths, but it is quite possible to add a new research instrument to the musicology laboratory.

^[1]           When analyzing the text by a computer program, the annotations in English accompanying the articles were excluded from the source material, since they are duplicated in the original language. The musical fragments offered by the authors in the form of illustrations are also not subject to analysis, since the examples are a graphic image.