Semantic Analysis of Group Values Structure Using Roget Thesaurus: Automated Algorithm

Abstract

The article proposes an approach for vectorization and quantitative analysis of group values. To demonstrate the possibilities of the method, the value structure of a group of young people was analyzed for differences between women and men, and for differences in the use of different parts of speech. The values of the group were verbalized in the form of free associations “with something most important in life”. The resulting array of words was converted into an array of semantic groups using Roget Thesaurus. Pairwise comparison of vectors with frequencies of individual semantic groups showed a high level of cosine similarity (0,9664) between subgroups separated by gender. Calculation of statistically significant differences in frequencies of separate semantic groups by сhi-square test allowed us to single out separate semantic groups, for which gender subgroups differed significantly. Frequency vectors obtained from the transformation of arrays of different parts of speech had a low level of cosine similarity in all pairwise comparisons. Nouns were most frequently used to express life values related to cause-and-effect relationships (14% of semantic groups). Adjectives were most often used to express values having a sense of personal predilections (18% of semantic groups). Verbs were most often used to express values related to liking (14% of semantic groups). The developed automatic algorithm will be useful for quantitative comparison of values between different groups, as well as calculating the degree of consistency of the target group values with the declared values of commercial brands.