The paper is devoted to the problems of automatic construction of the terminological system of the subject domain. A method for extracting domain terms from electronic encyclopedic data sources is proposed. The peculiarity of the proposed approach is a thorough analysis of the term structure, recognition of errors based on their linguistic classification, automatic generation of lexical-syntactic patterns representing multi-component terms, and the use of a set of heuristic methods for processing "special" terms. By analyzing encyclopedic dictionaries, a reference list of concept names is automatically formed, which is used to assess the quality of the dictionaries being developed.