The current stage of the tools development for processing Russian-language texts is associated with a weak elaboration of algorithms for identifying entities and dividing sentences into semantic, logically justified parts. In particular, how to determine an actor, actions performed by him/her, and the object(s) over which these actions are performed in a sentence. It is important to understand that the NLP algorithm should be designed in such a way that would identify the listed elements of the sentence not by the principle of a pre-compiled dictionary and the coincidence of parts of the sentence with it, but on the basis of highlighting universal dependencies. To this end, the team of authors have developed and tested algorithms based on the design of patterns universally describing various parts of a sentence: an actor, his/her actions and objects that these actions are directed to. To solve this problem, authors used the tools for parsing sentences into parts of speech and their dependencies (StanfordNLP); processing dependencies and identifying the tree structure of the sentence.