Expert assessment of synonymic rows in ruwordnet

Abstract

© Springer Nature Switzerland AG 2020. This article explores the principles of synsets in the RuWordNet thesaurus and synonyms in the classical dictionaries of Russian synonyms (N = 10) to identify discrepancies and improve the principles of organising synsets in RuWordNet. The relevance of the study is determined by the demand for WordNet resources in natural language processing tasks. The authors selected 102 RuWordNet thesaurus synsets, including nouns (N = 34), adjectives (N = 34) and verbs (N = 34). The meanings of the lexemes were correlated according to the data given in Russian language thesauri (N = 2). The comparative method and an independent expert assessment of RuWordNet revealed a number of discrepancies and inaccuracies in the representation of synsets concerning polysemy, hypo-hyperonymic relationships, lexical meanings of words and parts-of-speech synonymy. On the basis of this study, the authors recommend the elimination of individual shortcomings in the construction of the RuWord-Net synsets, in particular the polysemy and parts-of-speech synonymy.