Incomplete data analysis for building formal ontologies

The article considers the problem of automating the formation of ontological specifications subject domains on the basis of measurements. This problem is the pivotal issue of ontological analysis. The article presents models and methods, aimed at identifying conceptual structure and, ultimately, detecting the formal ontology of the considered subject domain. Fundamental realities of accumulation of empirical data (multiple independent measurements for each training sample properties of the object; congruence of the measurement procedures; differentiation of trust to different data sources) are reflected in the "objects-properties" summary table model. Imperfection (inaccuracy, inconsistency, uncertainty) of this information implies the need to use many-valued logic models for its primary processing. The result of this treatment - fuzzy formal context - should be approximated as a unique context, from which the possible conclusion of formal concepts in the framework of applied branch of the lattice theory, known as the "formal concept analysis". The genesis of the "properties' limits of existence" that affect the correctness of the approximation of fuzzy formal context is studied. The models and the method of accounting for this additional information are proposed. Guidelines for conversion of lattice formal concepts into a formal ontology are formulated. A model example of the developed models and ontological data analysis methods is presented.