The article provides two approaches for the implementation of the inference search procedure in the ontological base. One is based on the SWRL-rules, the other is a system with the natural language processing elements. The procedures have been established as a part of the recommendation system, which is developed at the Faculty of Informatics at National University of Kyiv-Mohyla Academy.We also add a description of the created approaches with their fragments of the taxonomic hierarchy for the planimetry ontology. For the performance examples, simple open-type problems with a short answer taken from the school geometry textbooks are used. The features of the approaches, how they work, as well as the capabilities they have are reviewed.The approach with natural language processing capabilities has a module for preprocessing raw Ukrainian text using the UDPipe 2.12 model, a module for rechecking the lemmas by using VESUM dictionary, a module with a described planimetry ontology, and a module for creating an illustration of the figures (triangles).To better illustrate the capabilities of the approaches on equal terms, we tried to use the same geometric problem. English translation of the problem: «Perimeter of an isosceles triangle = 40 cm and base = 10 cm. Find the legs of the triangle.». To solve this problem, systems should have a rule that subtracts the base from the perimeter, divides it by two, and sets the result to the correct variables (in our case, the legs of the triangle). We demonstrated that both approaches solved the problem successfully. But in order to achieve it, minor changes were added. Therefore, the more complete the ontology is, the greater the number of problem types the systems are able to cover.Having analyzed the results of the study, we can conclude that the systems are effective for solving geometric problems. The next step may be to combine the capabilities of the approaches to form a more complete knowledge base.

У статті наведено два приклади реалізації виводу в онтологічній базі знань. Один — із використанням SWRL-правил, другий — як систему з елементами обробки природної мови. Наведено опис створених фрагментів таксономічної ієрархії для предметної галузі (планіметрії). Як приклади використано прості задачі відкритого типу зі шкільного підручника з геометрії. Продемонстровані процедури є частиною рекомендаційної навчальної системи, яку розробляють на факультеті інформатики Національного університету «Києво-Могилянська академія».