NLP-BASED VERIFICATION OF MESSAGE RELIABILITY USING SEMANTIC NETWORK ANALYSIS

This article focuses on methods and approaches for constructing semantic networks for textual (news) messages in media streams to identify potential sources of disinformation. The main idea involves developing a comprehensive methodology for building such networks, where key terms are used as the foundation for semantic modeling. The authors analyze various text processing techniques, including preliminary computational text processing, extraction of key terms, and the identification of semantic relationships between them. Particular attention is given to the development of a metric for measuring semantic proximity between information messages represented as semantic networks. The proposed metric, based on the Frobenius norm, enables an effective evaluation of the similarity and interconnection between texts. This enhances the accuracy of semantic content analysis, uncovers hidden semantic relationships, and facilitates the structuring of information. Using the Frobenius-based metric, the article proposes an approach for identifying reliable and unreliable information sources, enabling further validation of the facts presented in news messages. This approach enhances the efficiency of information analysis, identifies trends, and predicts the development of events within the news space. Most importantly, it allows for detecting information influences, contributing not only to maintaining information security but also to ensuring national resilience against external threats. Keywords: semantic network, Frobenius measure, text analysis, Horizontal Visibility Graph algorithm, Directed Weighted Network of Terms, verification of message reliability.