FRACTAL PROPERTIES OF HUMAN LANGUAGE: FROM THEORETICAL MODEL TO EMPIRICAL ANALYSIS

Abstract

Background. The relevance of this study is driven by the growing interest of scholars in studying language from a synergetic perspective, where language is viewed as an open, dynamic, and non-linear system. One of the key concepts of this approach is fractality – the property of self-similar structures at different scaling levels. Recently, the application of fractal analysis has transitioned from a metaphorical to an empirical stage, particularly with the use of quantitative methods. Despite significant progress, a number of unresolved issues related to methodology and data interpretation remain. The aim of this article is a comprehensive analysis of the fractal properties of language based on the integration of principles from linguosynergetics, quantitative linguistics, and the empirical analysis of language data.

Methods. The research employs a combined approach, utilizing both qualitative and quantitative methods. The qualitative analysis involves the application of a fractal model of information deployment. The quantitative analysis is implemented through Detrended Fluctuation Analysis (DFA) and its multifractal generalization (MF-DFA).

Results. The qualitative analysis demonstrated that fractal patterns manifest at all levels of language: phonological, morphological, syntactic, and textual. The analysis of a media text confirmed that its structure is organized according to a fractal principle: key information is repeated at three scaling levels with increasing detail. The quantitative analysis of a sequence of sentence lengths using DFA revealed a scaling exponent of α=0.72, which significantly exceeds the 0.5 threshold and indicates the presence of strong long-range correlations and a non-linear text organization. The application of MF-DFA showed a wide multifractal spectrum, indicating a high heterogeneity of the text structure, where more ordered and informationally dense irregular fragments alternate.

Conclusions. The conducted research confirms that human language exhibits fractal properties at various levels of its organization. The combination of qualitative and quantitative analysis proves that text structure adheres to the principles of self-similarity, and the α-, φ-, ω-fractals model is an effective tool for describing the deployment of information. The practical significance of this work lies in the potential to use its results for improving NLP algorithms (automatic summarization), developing writing methodologies in language pedagogy, and identifying authorial style in discourse analysis. Prospects for further research include expanding the empirical base, conducting cross-linguistic comparisons, and analyzing large text corpora.