Прикладная и математическая лингвистика | Филологический аспект №09 (113) Сентябрь 2024

Структура мысли в поэтическом тексте

Тимофеева Мария Кирилловна
доктор филологических наук, доцент, ведущий научный сотрудник Федерального государственного бюджетного учреждения науки Институт математики им. С. Л. Соболева СО РАН, г. Новосибирск, Российская Федерация

Аннотация: Природа текста противоречива: форма его линейна, но выражаемая им мысль не линейна. Делинеаризация текста путём выявления его внутренней структуры и последующего разворачивания её в двухмерном пространстве открывает новые возможности восприятия и анализа текста. Цель статьи состоит в изучении этих возможностей на основе использования Теории риторических структур (ТРС) применительно к материалу русской классической поэзии (более 5000 словоупотреблений из стихотворного наследия А. К. Толстого и К. К. Случевского). Данная работа представляет первый проект по использованию ТРС для анализа русской поэзии. Полученные результаты демонстрируют применимость ТРС для сравнительного анализа структур мысли, присущих стихотворениям разных авторов и разным периодам их творчества. Представляемый проект открывает возможность обогащения цифровых вариантов стихотворных произведений путём дополнения их пространственными схемами, размеченными в терминах ТРС.
Ключевые слова: стиль рассуждения, внутренняя структура, стихотворный текст, сравнительный анализ, русская классическая поэзия

Timofeeva Mariya Kirillovna
Doctor of Philology, Docent, leading researcher at Sobolev Institute of Mathematics, Siberian Branch of the Russian Academy of Sciences, Novosibirsk, Russia, timof@math.nsc.ru

Abstract: The nature of text is controversial: it is linear by form although the expressed thought is not linear. Delinearization of a text by revealing its intrinsic structure and thus unwrapping the text into two-dimensional space opens new possibilities for perceiving and analyzing this text. The article aims to investigate and to assess these possibilities on the base of Rhetorical structure theory (RST) applied to the material of Russian classic poetry (more than 5000 word-usages written by A. K. Tolstoy and K. K. Sluchevsky). The current work presents the first project that attempts to use RST for analyzing Russian poetry. The obtained results prove availability of comparisons between the structures of thought intrinsic to the poetry of different authors and in different periods of their creative work. The project opens the way for enriching the digital counterpart of poetic treasures by supporting them with extensional models annotated in terms of RST.
Keywords: style of reasoning, intrinsic structure, poetic texts, comparative analysis, Russian classic poetry

1. Introduction

Contemporary culture of digitalization goes far beyond the task of making the digital twin of a considered work of art. Different modes of operating with digitalized counterparts are available. Present-day techniques and methodologies give rise to various tools for analytical processing, calculating, summarization, and navigating; and the set of tools renews rather rapidly. Consequently, a user gains possibility not only to get access to a digitized object per se, but also to study different strata of its representation with the help of convenient computer-aided instruments. We propose to enrich digital representation of texts by adding the stratum, which allows examining structures and styles of reasoning reflected in a text.

The way from the form of a text to its content passes through several associated levels, among them, phonetic, morphological, lexical, syntactic, semantic, pragmatic, and discursive. Only the first three of these levels are predominantly linear (albeit with certain variances). All other levels are apparently not linear. Pragmatic and discursive levels are the first and foremost objects of our concern, although semantics and syntax are also relevant.

Our main objective is to provide empirical support for benefits of analyzing and comparing structures of reasoning presented in digitized texts. The paper focuses on poetic texts however the proposed approach is applicable to prosaic texts as well.  

Human interest in studying strategies of reasoning has very long history, especially in its application to philosophic and scientific reasoning [1]. The contemporary methodology proposes several instruments for constructing such structures, among them the basic are argumentative analysis and rhetorical structure theory. Argumentative analysis is meant for presenting the schemes of argumentation, that is the combinations of arguments and theses that they support or attack; non-argumentative parts of a text are skipped [2]. This way of analysis misses the fact that besides the texts of predominantly argumentative nature (like scientific, philosophical, juridic) there are non-argumentative types that reflect not arguments and theses, but impressions, emotions or feelings, and present them in quite intelligential form. Coherence of a text may be achieved by non-argumentative relations as well as by argumentative. Poetic texts give good examples of this kind, and the Rhetorical Structure Theory (RST) [3] provides convenient tool for analyzing non-argumentative types of texts.

It is worth mentioning that the authors of RST, W. Mann and S. Thompson, as well as their adherents and colleagues, intended this theory for analyzing only prosaic texts. The history of RST summarized in [4] clearly shows the same destination as well as more recent surveys [5]. All the available applications of RST to Russian texts also work with prosaic texts [6, 7, 8]. Nevertheless, a poetic text, if it has some storyline (may be emotional of woven from impressions), can be the appropriate object of rhetorical analysis.   

The focused search for applications of RST to poetic texts produced only one result: that was the project devoted to Chinese narrative poetry [9]. Thus, the project presented in [10] was the first attempt to apply RST to Russian classic poetry. On this base, the subsequent project was initiated, it concerns structures of reasoning presented in song poetry [11].

The research method used in our project bases on RST and expands the applications of this theory by its adaptation to poetic texts. We used the instrument RSTTool (http://www.wagsoft.com/RSTTool/) for building the rhetorical structures.

In our investigation, we follow two lines of consideration. Firstly, we explore the modes of representing and analyzing RST-structures in dependence to the focus of our attention. Secondly, we argue for several investigative possibilities and illustrate them on the base of empirical work fulfilled on the material of Russian poetic texts. 

The structure of the subsequent text includes the following items. In the second section, we characterize the principles of applying RST to poetic texts and the basic parameters of the analyzed material. The third section presents the empirical results, and in the fourth section, we discuss and evaluate these results. The fifth section draws the conclusions of the current stage of investigation and outlines its perspectives.

2. Materials and Methods

The goals of the current investigation are:

1. To explore the structures of reasoning intrinsic to poetic texts and to judge about the potential of comparing the structures built for different classes of poetic works. Primarily, to compare the structures that correspond to the poetic texts written by different authors and in different periods of an author’s creative work. The delineated types of comparison have the potential of enriching stylistic analysis by pragmatic and discursive levels of consideration.
2. To formulate the prospective lines of further research.

The ordinary procedure of applying RST includes two steps. The first consists in dividing a text into elementary segments; each elementary segment designates the elementary thought, or “quantum of thought” [6]. The second step involves the process of determining rhetorical relations between the parts of a structure, namely between the elementary segments or the structures of these segments. The procedure results in building the tree diagram that represents the rhetoric structure of the text. The illustrative example is shown in Fig. 1. The whole structure is rather large; therefore, three substructures are collapsed into symbolic triangles, and each of them corresponds to the subtree of the whole tree diagram.

In general, both abovementioned steps of applying RST are rather complex. Nevertheless, in the case of poetic texts, the first step is relatively trivial because a text of this kind is already divided into lines, and it is naturally to consider this division as basic for segmentation. The aforementioned attempt of applying RST to the Chinese poetic texts [9] follows the same strategy of segmentation. Besides, the borders between lines in a poetic text often mark the places of speech prosody, and prosody is one of the important reasons for demarcating the segments [6].

The only exception was made for the lines that include a sign of a dot (dots), question mark, or mark of exclamation in internal positions of a line (not as the last sign of a line). In such a case, every sign of the mentioned category indicates the borderline between segments; thereby transferring the line into concatenation of the elementary segments.

Figure 1. The rhetorical structure for the poem “Song” written by A. K. Tolstoy.  Translated in English by Lindsay S. Perkins [13].

The second step is substantially more complicated than in general because a poetic text rarely includes explicit and unambiguous textual markers of rhetorical relations. Thus, it was necessary to elaborate the special situational and logical principles for identifying rhetorical relations.

For making our research, we used the set of 34 sample relations, among them 27 classical types that were defined by the authors of RST, and 7 types that we introduced in our research.

Following the initial taxonomy of W. Mann and S. Thompson, we divide the whole set of rhetorical relations into two classes: presentational relations and subject matter relations. The difference between these classes bases on an author’s intention. By using a presentational relation, an author intends to increase some inclination in a reader, to influence a reader’s behaviour, attitudes or thoughts. A subject matter relation corresponds to another type of intention. This intention is informational: an author wants to inform a reader about some state of affairs; as a result, the reader should recognize the relation in question.

The basic empirical material of the current work includes the lyrics of A. K. Tolstoy (30 texts, or 3455 word-usages) and the lyrics of K. K. Sluchesky (30 texts, or 2473 word-usages). All the considered texts are in Russian. We compared two periods of K. K. Sluchesky’s creative work (1856-1883 and 1895-1901). We also drew a comparison between the late creative periods of A. K. Tolstoy and K. K. Sluchesky. Besides, we compared the whole set of structures built for the poetic texts written by A. K. Tolstoy with the whole set of structures built for the poetic texts written by K. K. Sluchesky. For exploring perspectives, we considered some translations of poetic texts.

For every line of comparison, we contrasted the quantitative characteristics of rhetorical relations, namely frequencies of their usage. In addition, we considered several classes of these relations and compared their frequencies in the analyzed types of poetic texts.   

3. Results

The basic result of the work consists in building the corpus of poetic texts annotated in terms of RST. This variant of digitalization opened the way to several sorts of further investigations.

Primarily, the corpus permits focusing on different aspects of a rhetorical structure. Fig. 1 gives the illustration. Since the whole structure of the text is rather large and does not fit into a page size, three substructures are collapsed into triangles. Nevertheless, the depicted image permits us to see general organization of the author’s reasoning. Over against, if we intend to focus on a certain local substructure, it is possible to explicate it in all details and collapse into triangles the other substructures. This variability is important because usually the structures are rather space occupying, and it is essential to study the trajectory of thought both in whole and in detail.

The explored lines of compatison have demonstrated other possibilities of representing a poetic text. Each text may be characterized by the set of its rhetorical relations. Consequently, by summarizing the corresponding statistics we can juxtapose frequences in different classes on texts. For each specific relation, we have calculated the frequency as  the per сent of  the whole amount of relations encountered in the structures that were built for the class of the verses under consideration.

For instance, Fig. 2 shows frequency ratio for rhetorical relations in the texts written by A. K. Tolstoy and the texts written by A. K. Sluchevsky. 

Several types of relations (among them conjunctions) were excluded from consideration because they are very frequent in all analysed texts and therefore are not informative for comparisons.

Figure 2. The frequencies of rhetorical relations in the lyrics of two authors: A. K. Tolstoy (red lines) and K. K. Sluchevsky (blue lines).

As we can see on the Fig. 2, three types of relations – Evaluation, Contrast, and Evidence – are much more characteristic for the verses written by K. K. Sluchevsky than for the verses written by A. K. Tolstoy.

On the contrary, there are no substantive differences between the frequencies of relations for the basic and enlarged samples of K. K. Sluchevsky’s verses from the late period (Fig. 3). The basic sample includes 30 verses (2473 word-usages), the enlarged sample includes the basic sample and 10 additional poetic texts (774 additional word-usages).

Figure 3. The frequencies of rhetorical relations for the basic and enlarged samples of the K. K. Sluchevsky’s verses from the late period (the red lines correspond to the basic sample, the grey lines – to the enlarged sample).

One more line of comparison consists in aligning and collating different translations of the same rhyme into another language. This juxtaposition permits us to elicit invariant and varying parts of the structures. For instance, Fig. 4 represents the juxtaposition of six translations made be different authors for the Ch. Baudelaire’s rhyme “Correspondences” [14]. As we can see, two types of relations (Contrast and Comparison) are invariantly present in all translations. Other rhetorical relations are varying between the translations. The diagram makes available a very user-friendly way of presenting the situation.  

Figure 4. Collating six translations of the Ch. Baudelaire’s rhyme into Russian.

In addition to the detailed comparisons like these depicted in the Fig. 2, Fig. 3 and Fig. 4, there can be more general comparisons based on different taxonomies of the relations. For instance, in [10] the integral characteristics were elaborated and illustrated by the series of comparisons between three periods of A. K. Tolstoy’s creative work, including his lyrics, prose, and translations of the other author’s poetry. 

Table 1 presents the results of highly general taxonomy that divides the whole set of rhetorical relations into two classes: presentational relations and subject matter relations. For this comparison, we divided the poetry of A. K. Tolstoy into two periods: 1840-1858, 1859-1854. For K. K. Sluchevsky the corresponding division is more distinct because there is more than ten-years break in his poetic work. Thus, for K. K. Sluchevsky the periods are 1856-1883, 1895-1901.

Table 1. Correlation between presentational and subject matter relations.

Tab. 1 gives evidence for rather stable difference between frequencies of presentational and subject matter relations. This gives rise to hypothesis about invariant character of this ratio for Russian poetry, although the assumption needs verification on more representative collection of poetic texts, including the poetry of other authors.   

One more potentially invariant feature of poetic texts consists in regularities of conjunctions. Frequency of conjunctions is characteristic to all the analyzed poetic texts. For instance, Fig. 1 visualize the conjunction with four conjuncts. Poetic texts are rhythmical, and this feature is intrinsic not only to its form but to its content as well. Repeated and nesting conjunctions are rather frequent in poetry and they are good candidates for filling poetic thought with rhythmicality. Very often, a conjunction has one or two first simple conjuncts and the final conjunct that is more structurally complex. Other types of relations and their combinations also may display rhythmicality of thought. These regularities need further more focused examination.

4. Discussion

Investigations in the area of individual author’s poetic styles usually go along the traditional linguistic lines, namely, they uncover these phonological, lexical, morphological, and syntactic peculiarities that are congenial to the author. For instance, Iu. M. Lotman [12], the well-known specialist in semiotics and poetry, considers all the above-mentioned aspects. Besides, he studies peculiarities of a poetic plot, as well as the specific poetic features like rhythm and cadency. Although he does not consider trajectories of thought that are characteristic for poetry in whole or for an author’s style in particular.

Our research proposes the method and the preliminaries for studying the way of filling the gap. The explored approach gives the results of two kinds.     

First, it proposes new hypothesis concerning poetic texts and their peculiarities. The revealed regularities permit to put forward several new supposals about styles of reasoning characteristic to A. K. Tolstoy and K. K. Sluchevsky. For testing these supposals, it is necessary to continue investigations on the material of other poets. This subsequent analysis of RST structures should go along two lines: comparing the structures of reasoning intrinsic to other authors, and tracing the changes of these structures with time.

Second, along with formulating new hypotheses, the explored approach can give additional arguments for already existing hypothesis. More specifically, the entertained investigation gives support for the V. Bryusov’s judgement about the poetry of K. K. Sluchevsky.  V. Bryusov characterized K. K. Sluchevsky as the poet whose verses are full of contradictions, confrontations, and dissonances [15, с. 231-234]. As we can see in Fig. 2 and Fig. 3, the revealed structures of thought give evidence for this judgment and demonstrate that this feature is much more characteristic for K. K. Sluchevsky than for A. K. Tolstoy. We see, for instance, that both poets use the relation of Contrast, although K. K. Sluchevsky do it more actively. Whereas V. Bryusov made his supposition introspectively, RST tool gives us opportunity to study this peculiarity in details, to consider what sorts of confrontations are intrinsic for the poetry of K. K. Sluchevsky and what their quantitive characteristics are.

Additionally, the structures of thought represented in terms of RST provide us with convenient tool for focusing on the details of a structure that are actual at present. It is possible to collapse the irrelevant substructures and expand the relevant ones. Besides, RST provides us with possibility of visualizing and juxtaposing the schemas of reasoning presented in the parallel translations.

5. Conclusion

The most obvious benefit of digitalization consists in accessibility and conveniency of navigating through the selected layers of information concerning an object of art. Benefits that are more specific add options for considering and studying different strata of the object’s representation. It is especially important for the objects of art because they usually have complex semiotic organization. In particular, structures of reasoning are essential for textual data.   

The obtained results propose positive support to the main issue of the research: they show that RST is applicable to poetic texts and can reveal regularities in an author’s style of reasoning as well as specific features of poetic reasoning that are common to poetry in whole or to the specific groups of authors. This aspect of research in novel and has the potential of enriching stylistic investigations.

Besides, it can contribute to the contemporary knowledge about human reasoning in whole. This is particularly important for the corresponding areas of natural language processing and artificial intelligence. Moreover, the resulting corpus of poetic texts annotated in terms of RST can serve as the additional data for machine learning. 

Acknowledgments

The study was carried out within the framework of the state contract of the Sobolev Institute of Mathematics (project no. FWNF-2022-0012).

1. Sciortino L. What is a Style of Reasoning? Transversal: International Journal for the Historiography of Science, 2023. V. 15. P. 1-20.
2. Sidorova EA, Akhmadeeva IR, Zagorulko YuA, Kononenko IS, Sery AS, Chagina PM, Shestakov VK. An integrated approach to the analysis of argumentative relationships in scientific communication texts // Ontology of designing. 2023. V. 13(4). P. 562-579. [In Russian]
3. Mann W., Thompson S. Rhetorical Structure Theory: Toward a functional theory of text organization. // Text – Interdisciplinary Journal for the Study of Discourse, 1988. V. 8 (3). P. 243-281.
4. Taboada M., Mann W. Applications of Rhetorical Structure Theory // Discourse Studies, 2006. V. 8(4). P. 567-58.
5. Hou S., Zhang S., Fei C. Rhetorical structure theory: A comprehensive review of theory, parsing methods and applications // Expert Systems with Applications, Volume 157, 2020, 113421.
6. Kibrik A. A. & Podlesskaya V. I. (Eds.) Night Dream Stories: A corpus study of spoken Russian discourse. 2009. Moscow: Languages of Slavonic Culture. 736 p.
7. Toldova S., Davydova T., Kobozeva M., Pisarevskaya D. Discourse features of blogs in subcorpus of Russian Ru-RSTreebank // Computational Linguistics and Intellectual Technologies: Proceedings of the International Conference “Dialogue 2020”. 2020. С. 747-761.
8. Смирнов И.В. Интеллектуальный анализ текстов на основе методов разноуровневой обработки естественного языка. М.: ФИЦ ИУ РАН, 2023. 356 с.
9. Lu Y. Applying Rhetorical Structure Theory to classical Chinese poetry: a case study of Mulan Ci. In: Proceedings of the 2nd ASEAN Plus Three Graduate Research Congress (2nd AGRC). Bangkok, Thailand, Mahidol University, 2014, p. 415–423.
10. Timofeeva, М.К. The Rhetorical Structure Theory as an Instrument for Analyzing Poetic Texts // Tomsk State University Journal of Philology, 2020. V. 68. P. 109–137.
11. Zorkaltsev A. V. Rhetorical Structure Theory: Analysis of Aesthetics and Complexity of Song Poetry // 2024 IEEE 25th International Conference of Young Professionals in Electron Devices and Materials (EDM). Altai, Russian Federation, 2024. P. 2210-2213.
12. Lotman Iu. M. Analysis of a poetic text: The structure of poetry. Leningrad: Russian Education Holding “Prosveshcheniye”, 1972. 309 p.

Список источников
13. Perkins L. S. Song. By Count Alexis Tolstoy // The Russian Review: A Monthly Magazine, Devoted to Russian Life, Literature and Art, 1916. V. 1(4). 207 p.
14. Shamanskij D. Correspondences (Several translations of one Сh. Baudelaire’s rhyme). Russian Language Abroad, 2002. V. 2. P. 98-103.
15. Bryusov V. Y. The poet of contradiction. In: P. Antokol'skij, A. S. Myasnikov, S. S. Narovchatov, & N. S. Tihonov (Eds.) // Bryusov V Collected works in seven volumes. Vol. VI. Moscow: State-Publishing House of Fiction1975.