The main purpose of natural language processing is to provide a computer with the ability to understand natural language orally and in writing, as well as to produce responsive results. The difficulties in developing such systems are related to the fact that natural language is full of ambiguities, which make it difficult to write software that generates grammatically and semantically correct sentences in natural language. This paper describes a mechanism for inferring information from intelligent knowledge bases using the semantic role labeling method. This method makes it possible to determine the semantic and syntactic structure of the inferred sentence. It is shown that for the effective work of the inference mechanism it is necessary to have a large amount of semantically labeled data. The concept of the designed program for implementing semantic role labeling is described. The results of semantic role labeling of aviation and astronautics texts are presented. A unique inventory of semantic roles for the aerospace texts has been compiled. This can be used as a basis for the development of systems requiring automatic semantic analysis.