ҚАЗАҚ ТІЛІНДЕГІ ЭКСТРЕМИСТІК ИДЕОЛОГИЯНЫ АНЫҚТАУ: АННОТАЦИЯЛАУ МӘСЕЛЕЛЕРІ ЖӘНЕ ТЕРЕҢ ОҚЫТУ ТӘСІЛДЕРІ

Abstract

Abstract. Destructive content currently distributed on social media platforms and instant messengers poses a serious threat to information security. The increase in the volume of content and the complexity of the methods of its hidden transmission reinforce the need for automatic detection of such materials. The lack of funds and methods, especially in the field of analyzing texts in the Kazakh language, makes this problem urgent. The main objective of the study is to establish a methodological framework for the automated analysis of extremist content in the Kazakh language and to systematize the capabilities of existing approaches. The paper analyzes the main ideological categories of extremist content – propaganda, recruitment and radicalization – and sets the task of classifying the text in these areas. The study provides for the compilation of a balanced data collection based on materials in the Kazakh language collected from the VK, YouTube and Telegram platforms. When annotating, it is noted that differences in expert opinions, the possibility of using soft labels and disparagement-based learning approaches are taken into account. Annotator disagreements are treated as informative signals, and three disagreement-based models — MO-WEL-4xBERT, MO-WEL-4xLLM, and MO-WEL-Classic-Ensemble — are compared. The best result was achieved by MO-WEL-Classic-Ensemble (F1=0.9664), demonstrating the advantage of heterogeneous ensemble methods. Classical models and deep learning models are used to automatically identify extremist content. The effect of text preprocessing on the accuracy and accuracy of the model is also shown. The article offers a comprehensive methodology for analyzing the content of social networks in the Kazakh language and identifying extremist ideology using artificial intelligence and natural language processing technologies. The proposed approach is aimed at strengthening information security, as well as ensuring ideological stability in society.