FORMALIZATION OF RECOGNITION AND IDENTIFICATION OF SEMANTIC OBJECTS IN NATURAL LANGUAGE TEXT STREAMS

Abstract

The increasing incidence of crimes committed in cyberspace, particularly on social networks and various messengers, necessitates the development of adequate and effective countermeasures. The rise in cybercrime is so significant that it poses a potential threat of inflicting irreparable harm to the state and society. However, detecting such crimes and criminal activities is challenging because offenders operate virtually and linguistically within social networks, exploiting their features to conceal their traces. Nonetheless, various detection and identification tools capable of automatically processing natural language, highlighting specific semantic features of criminal activities, and recognizing and identifying them could serve as effective countermeasures. Given the impracticality of applying neural network approaches to these situations for several reasons, this study proposes a formal method for designing a recognizer to identify semantic objects in text streams based on their linguistic traces. Formal concepts such as the formal model of a semantic object, behavior function, scenario, linguistic trace, and recognition function are introduced. The reasoning is based on set-theoretical principles of computational theory of semantic interpretation and utilizes computational representations of the meaning of text fragments for their comparison in terms of semantic similarity. The proposed approach is general and universal, allowing for the formal synthesis of a recognizer for semantic objects based on their linguistic descriptions and behavior. All discussions and constructions in the work are illustrated with specific examples.