Background. Development of control tests of learning outcomes requires creation of test question. The number of these questions should be sufficient for reliable knowledge assessment, and their content should comprehensively reflect knowledge in the subject area of control. Currently, methods of automatic question generation based on machine learning models and rational models of knowledge representation are used. Their application completely removes the quantitative component of the problem. However, machine learning models do not guarantee a holistic representation of knowledge, and rational ones, although they provide integrity, do not solve the problem of generating questions that control understanding of the simple rules necessary to obtain a given result. The purpose of the study is to solve this problem.
Materials and methods. The rational model of subject knowledge is presented according to the web ontology language standard. This choice is justified due to wide use of the standard, the availability of accessible documentation and freely distributed applications, in particular, the Protégé ontology editor and the SWI Prolog programming environment, which declarative knowledge representation model is compatible with the web ontology model.
Results. A method for synthesizing control questions based on a model of subject knowledge has been developed. Synthesis is performed by inversion of axioms of the following form: “if the properties of an entity satisfy the constraints of an OWL class, it belongs to this class.” The result of the inversion is a question of the following form: “what constraints should the properties of an entity satisfy if it belongs to an OWL class?” Inversion is a feature of the method, which, unlike the known ones that generate questions for understanding the consequences of known causes, synthesizes questions for understanding the causes of a given consequence. The validity of results is confirmed by their comparison with the “axiom-consequence” pairs of the ontological model.
Conclusions. The method ensures control of understanding of the rules necessary to obtain a given solution using the axioms of OWL classes. It has been used to develop tests for engineering training courses. The use of a standard format for knowledge representation allows the method to be used for developing control questions in other subject domains.