The study of the statistical characteristics of the text based on the graph model of the linguistic corpus

E. G. Grigorieva ; V. A. Klyachin

Izvestiya of Saratov University. Mathematics. Mechanics. Informatics, Tome 20 (2020) no. 1, pp. 116-126

Résumé

The article is devoted to the study of the statistical characteristics of the text, which are calculated on the basis of the graph model of the text from the linguistic corpus. The introduction describes the relevance of the statistical analysis of the texts and some of the tasks solved using such an analysis. The graph model of the text proposed in the article is constructed as a graph in the vertices of which the words of the text are located, and the edges of the graph reflect the fact that two words fall into any part of the text, for example, in — a sentence. For the vertices and edges of the graph, the article introduces the concept of weight as a value from some additive semigroup. Formulas for calculating a graph and its weights are proved for text concatenation. Based on the proposed model, calculations are implemented in the Python programming language. For an experimental study of statistical characteristics, 24 values are distinguished, which are expressed in terms of the weights of the vertices, edges of the graph, as well as other characteristics of the graph, for example, the degrees of its vertices. It should be noted that the purpose of numerical experiments is to squeak in the characteristics of the text, with which you can determine whether the text is man-made or randomly generated. The article proposes one of the possible such algorithms, which generates random text using some other text created by man as a template. In this case, the sequence of parts of speech in an auxiliary text alternation is preserved in the random text. It turns out that the required conditions are satisfied by the median value of the ratio of the text graph edge weight value to the number of sentences in the text.