Semantic etalons and knowledge transfer in the problem of their estimation based on the open type tests.

The article is devoted to the problem of knowledge transfer between experts and learners in machine learning and knowledge control systems that store information in the form of Natural Language (NL) text units. The purpose of this study is to minimize losses of useful information when forming a knowledge-based system that works with a textual description of the subject area test facts. The solution of this problem is suggested within the framework of the theory of Formal Concept Analysis (FCA) based on the concepts of Situations of Language Use (SLU) as a unit of formal description of the semantics. In the article, in particular, coordination of knowledge generated by experts as well as search for the most efficient transfer method of information between the two groups of NL carriers (experts and trainees) are considered to be very important tasks. In accordance to the model proposed be the authors of this article, the use of SLU etalons as units of the thesaurus and concordance of the etalons allows to reduce the size of the text data. The authors describe a system that performs a search of the SE-form closest to a user response, which defines SLU of the correct answer. Next comes the analysis of the word discrepancies, searching consistencies among mismatched responses’ parts being compared as a part of correct answer’s etalon and evaluation according to found synonyms. For use of such assessments in evaluating the expert knowledge from different industries, it was necessary to reformulate the definition of SLU similarity using fuzzy logic. System analysis of the professional knowledge structure in a particular area is used for a description of the membership functions.