SUMMARY
The task of evaluation of measurement uncertainty meaning of natural language constructs (NLC) based on formalization of the conceptsof linguistic image, artificial cognitive systems and unit of sense is resolved. The basis of model the knowledge base of artificial cognitivesystem laid down statistical information regarding the associative compatibility of linguistic images, which enables unified evaluation the unit and the quantity of sense NLC. The method for measuring the sense of NLC based on fuzzy relation of meaning is offered. It provides to use information about the links between lemmas of text that allows you to estimate the measurement uncertainty of two types of sense signs. The results of the formal evaluation of the uncertainty of measurement sense of NLC are received and interpreted what enables into account information about the relationship between lemmas for solve tasks of computational linguistics. With developed on the basis of the package DKPro Core Software conducted experiments to study the proposed method in the problem of the definition of informative features of the text. The experiments obtained dependence of the parameters detected Pareto-like distribution law relations between lemmas, whose analysis suggests that average number of connections of linguistic image is the most informative numerical feature for the text.