AUTOMATIC DETECTION SYSTEM OF UKRAINIAN-LANGUAGE DISINFORMATION BASED ON MACHINE LEARNING

Abstract

The development of tools for identifying and analyzing information threats is an urgent task that is of great importance for ensuring the information security of Ukraine, especially at the present time. Research and analysis of methods and complex tools for identifying fakes and disinformation not only meets the critical need of Ukrainian society for reliable means of information verification, but also improves the general culture of information consumption, strengthening the information resilience of the nation.

The article analyzes existing approaches and tools for identifying, assessing and countering disinformation, fake news and propaganda in the Ukrainian information space. A prototype of a text analysis system capable of potentially detecting manipulative fragments has been developed. The capabilities of the applied machine learning technologies for creating an appropriate model of the system, as well as the functional capabilities of the developed system, are described.

The article analyzes existing approaches and tools for identifying, assessing and countering disinformation, fake news and propaganda in the Ukrainian information space. A prototype of a text analysis system capable of potentially detecting disinformation using vectorization based on TF-IDF and a multinomial naive Bayes classifier has been developed. The scientific novelty lies in the application of these machine learning methods to Ukrainian-language content, as well as the use of FAISS for fast search of nearest neighbors and clustering in vector space. The results of comparing the model accuracy estimates for true and false content are presented. This allows us to determine for which type of texts the model works better.

Further research will be aimed at learning and training the model on new data, testing and evaluating the proposed system, as well as using this system for automated monitoring of news and in social media to identify potentially falsified information.