PROPERTIES OF AUTOSEMANTIC WORD NETWORKS IN UKRAINIAN TEXTS

Abstract

We present results of network analysis of Ukrainian texts. Autosemantic (meaningful) words are considered as network vertices connected with links when belonging to one sentence. Subnetworks corresponding to specific parts of speech (verbs, nouns, adjectives, etc.) are also built. The obtained networks are small-world and scale-free. To make comparisons, random texts with parameters corresponding to real texts are generated using several approaches. Various parameters of networks are calculated, including transitivity, betweenness, degree centralization, mean distance, network diameter, exponents of degree distribution, etc. Comparison of network parameters of real and generated texts shows that borders between them are quite fuzzy.