On determining semantic similarity based on relationships of a combined thesaurus

Abstract

Problems of the use of thesauruses for fuzzy comparisons of conceptual patterns are considered. A measure of semantic similarity that can be calculated using hierarchical and association relationships of a thesaurus is proposed, as well as an algorithm to compile a semantic intersection of conceptual patterns based on the coinciding maximum principle. A massive of texts and conceptual search patterns of thesis papers was used for experimental studies, which proved that the use of the lexis of different subject fields of a multi-area thesaurus produced a more precise identification of sematic similarity. The power of the pattern intersection increased significantly through pairs of descriptors linked by the semantic similarity measure; however, the average degree of pairwise intersection only increased by 1–2%, which implies an insignificant “expansion” of a conceptual pattern as it is used as a search pattern in creating search-result outputs in automated search mechanisms.