New metric that uses a measure of resemblance between terms to take into account the notion of semantic proximity


Zhaxybayev D.O. Bakiyev M.N.
30 March 2021Little Lion Scientific

Journal of Theoretical and Applied Information Technology
2021#99Issue 81915 - 1930 pp.

In this article, we extended the vector model by adapting the parameter by combining it with the formula for index word extraction and evaluation in order to describe the relevant principles that describe a text. Indeed, by combining the calculation with an approach, we have proposed a new metric that uses a measure of resemblance between terms to take into account the notion of semantic proximity. This indexation approach is supported by a contextual and semantic appraisal. In order to have a comprehensive descriptor index, we used not only a semantic graph to illustrate the semantic relationships between words, but also an auxiliary dictionary to strengthen the cohesion of the established graph and thus the semantic weight of indexation phrases. In the presented article, two semantic similarity approaches were explored in Kazakh-Russian, namely, the direct path-based and distributional model, and their cross-lingual counterparts were synthesized in the light of English. The suggested approaches were evaluated on a specific dataset of 1000 Russian and Kazakh word pairs, formatted by analysis. The correlation scores obtained between the four tests and the human evaluation scores suggest a major shift that brings the cross-lingual approach to the semantic similarity estimation process in the Kazakh and Russian languages.

Algorithm-based search , Automized search engine , Semantic-Lexical groups , Similarity , Verbal word identification and indexation

Text of the article Перейти на текст статьи

L.N.Gumilyov Eurasian National University, Department of Information Systems, Nur-Sultan, Kazakhstan

L.N.Gumilyov Eurasian National University

10 лет помогаем публиковать статьи Международный издатель

Книга Публикация научной статьи Волощук 2026 Book Publication of a scientific article 2026