Publications Repository - Gdańsk University of Technology

The methodology of semantic clustering analysis of customer’s text-opinions collection is developed. The author's version of the mathematical models of formalization and practical realization of short textual messages semantic clustering procedure is proposed, based on the customer’s text-opinions collection Latent Semantic Analysis knowledge extracting method. An algorithm for semantic clustering of the text-opinions is developed, the distinctive characteristics of which is the introduction of concepts and methods of identification point of reference in the scale of text-opinions collection closeness determination; instrument of the documents’ closeness degree identification; measure of similarity between pairs of documents. The version of quantitative evaluation of the clustering results is developed. The concepts of resolving power of the method of semantic clustering and level of the clustering procedure quality are proposed. Analysis of the specific features and the effectiveness level of various distance measures is conducted

Authors

• dr Nina Rizun link open in new tab ,
• dr inż. Paweł Kapłański link open in new tab ,
• professor Yurii Taranenko