Аннотация
Abstract In recent years, there has been a noticeable increase in both individuals and organizations utilizing social networks for illicit purposes. This trend can be viewed as a potential threat to the national security of the country. In this article, the authors pay attention to how various extremist organizations use social networks in their activities, and offer LSTM‐based models for classifying extremist texts in Kazakh on web resources. The main purpose of the article is to classify Kazakh texts in social networks into extremist and non‐extremist classes. The authors employed techniques such as Tf‐Idf, Word2Vec, Bag of Words (BoW), and n‐grams in experiments. A list of extremist keywords in the Kazakh language and, accordingly, a corpus of extremist texts in the Kazakh language were created for training and testing machine learning methods. As a result, the authors introduced a model that demonstrated superior performance across all evaluation metrics in machine learning for detecting extremist texts in the Kazakh language. The theoretical significance of this study lies in its comprehensive exploration of methods and algorithms for detecting extremist activities and organizations. The foundational findings derived from this research can contribute valuable insights to the global scientific community. The practical implications, including the developed methodology can be utilized by authorized entities to enhance information security, safeguard critical infrastructure, and combat online extremism.