Обоснована актуальность автоматизации создания лингвистических онтологий и выбор для этой цели Википедии в качестве источника информации. Предложена математическая модель Википедии и лингвистической онтологии, использующая парадигму нечетких семантических отношений между понятиями. Сделан краткий обзор мер семантической близости понятий с учетом их вычислительной сложности, обоснован выбор взвешенной меры Дайса.
The application without knowledge of an ontological type allows updating considerably quality of problem solutions in natural language processing. A number of researchers use Wikipedia as a basis for the formation of such resources. This paper reports the formalization method of Wikipedia structures and linguistic ontology used in the developed by the authors system of the linguistic ontology formation a specified subject field from Wikipedia. The papers and references connecting them serve a purpose for formation of a weighted graph of ontology to the graph nodes correspond notions, and to the ribs of graph – fuzzy semantic relations between them. The references obtain different weights depending on entering this or that information unit on a page. By a graph of relations it is possible to estimate numerically the degree of semantic proximity of two arbitrary concepts. For this purpose it is possible to use different measures of semantic proximity. Recursive measures possess considerable computational complexity at insignificant improvement of quality in test problem solution in comparison with nonrecursive local measures of the Dice measure type that is unacceptable for the ontology large enough. From these considerations the Dice weighted measure is chosen as a basic one for the system under development.