NEURAL NETWORK IDENTIFICATION OF PROPAGANDA GOALS IN TEXT CONTENT WITH VISUAL ANALYTICS OF RESULTS

Abstract

The work aim is to create a method for neural network identification of propaganda goals in text content with visual analytics of the results, which should include expanding the set of propaganda goals and using context windows to analyze the relationships between propaganda techniques and goals. Methodology. The developed method involves analyzing text content using neural networks, which allows identifying propaganda goals and techniques. Context windows are used to assess the relationships between propaganda techniques and goals. The results are presented in the form of visualization of the semantic model of propaganda, which includes assessing the semantic significance and importance of relationships. Scientific novelty. The main difference of the proposed method is the expansion of the set of propaganda goals by supplementing the set of their verbal variants. The introduction of context windows to analyze the relationships between propaganda techniques and goals allows to increase the accuracy of propaganda detection and provide a visual display of the results. Conclusions. Experimental results confirmed the effectiveness of the developed method, which demonstrates improved detection of propaganda targets compared to existing approaches. High correlation with expert assessments confirms the reliability and accuracy of the approach. Visual analytics provides transparency and explainability of the results, which contributes to a deeper understanding of the mechanisms of propaganda influence. The method has significant potential for use in automatic analysis of text content and can be adapted for various text sources.