Properties of the network of semantic relations in the Russian language based on the RuWordNet data

Abstract

It was previously shown that the WordNet thesaurus has a small-world structure. We obtained a similar result for RuWordNet, a recently created thesaurus of the Russian language, and determined the main characteristics of the network of semantic relations in the Russian language. They are the average length of the path between the vertices, the maximum path length between the vertices, the cluster coefficient and the assortativity coefficient. The distribution of the vertices degrees was also considered. We compared the values of the structural parameters of the networks for WordNet and RuWordNet. A significant difference between RuWordNet and WordNet is that RuWordNet has many semantic relations of the ‘POS-synonymy’ type (synonymy between words of different parts of speech) that are hardly found in WordNet. The values of structural constants were determined for RuWordNet both regarding this type of relations and without taking into account these relations, which allows one to estimate the influence of relations between parts of speech on the structure of the words network.