Sergey Soloviev– Professor, Department of Programming Languages, Faculty of Computational Mathematics and Cybernetics, Lomonosov Moscow State University.
Address: 1, build. 52, Lomonosov Moscow State University, Leninskie Gory, GSP-1, Moscow, 119991, Russian Federation.
E-mail: soloviev@glossary.ru

Daria Stelmashenko – Engineer, Faculty of Computational Mathematics and Cybernetics, Lomonosov Moscow State University.
Address: 1, build. 52, Lomonosov Moscow State University, Leninskie Gory, GSP-1, Moscow, 119991, Russian Federation.
E-mail: dashanikolaeva@gmail.com

The article looks into the problem of merging local contexts, describe the same global set of objects in the terms of different sets of attributes. The resulting solution of this problem is a formal concept lattice that is defined on a global set of objects via the unified set of attributes. Among the known methods for solving this problem, the @FC method has been chosen, which allows to precisely building the desired lattice. It is pointed out that the @FC method brings the solution of the problem to a sequence of formally-grounded classification solutions, on the basis of the partially given descriptions of objects. The complexity of solving the problem with @FC method is in the direct dependence from the number of successful classification decisions, which brings in the necessity of expanding their range. In similar conditions, an expert classification method has already performed as a rather effective instrument that allows transferring the once obtained classification decisions onto other objects. The base of the expert classification method is formed by the domination relation on objects descriptions, which is obtained during the dialog with an expert. The fundamental difficulty in incorporating of expert classification method into the @FC method is in the necessity of a self-consistent extrapolation of the domination relation on partial descriptions of objects from the global set. The article describes a concrete method for extrapolating the domination relation, which eliminates this difficulty and allows combining of these two methods. The immediate practical result of this research is the development of an effective modification of the context merging method. From the theoretical point of view, the proposed way of combining the two methods generates a number of interesting research topics.