Search academic texts

Information × Registration Number 0226U002577, (0125U002254) , R & D reports Title Linguistic expert analysis of Ukrainian-language destructive Internet content using artificial intelligence technologies popup.stage_title Створення лінгвістичного забезпечення системи TextAttributor 2.0 Head Darchuk Nataliia P., к.філол.н. Registration Date 27-02-2026 Organization Taras Shevchenko National University of Kyiv popup.description1 Development of an intelligent interactive system for automatic content analysis of Ukrainian-language destructive media texts, TextAttributor 2.0, based on artificial intelligence technologies (deep machine learning) and classical computer linguistics (rule-based method), which will monitor the Internet space and perform linguistic expert analysis in the task of identifying toxic and manipulative texts, as well as idiom and authorship popup.description2 The relevance of the project is determined by the need to counteract destructive information influence in the context of the Russian-Ukrainian war. The goal is to develop the TextAttributor 2.0 interactive intellectual system for automatic content analysis of destructive Ukrainian-language media texts based on artificial intelligence technologies and classical computational linguistics. The system performs the functions of monitoring the Internet space and linguistic expert analytics in the task of identifying toxic and manipulative texts, idiolect and authorship. The main tasks include: forming annotated datasets of Ukrainian-language destructive media texts; improving the TextAttributor 1.0 system; expanding the lexicographic database with destructive linguistic units; improving the statistical apparatus in stylometry and attribution tasks; developing modules for automatic semantic-statistical and syntactic-statistical analysis; developing a module for parameterising acoustic speech information; developing neural networks; creating software and a new system interface. The study uses a comprehensive interdisciplinary approach, which is reflected in the following working hypotheses: 1) the use of dictionaries and rules in morphological and syntactic analysis tasks, semantic identification of destructive means, statistical parameterisation of text will ensure the implementation of linguistic expert analytics regarding the presence of toxic and manipulative means and their categories in the text; 2) the use of a morphological-syntactic-semantic model of statistical analysis will allow determining the stylistic features of media text and the degree of similarity between two or more texts in tasks of establishing authorship; 3) the application of transformer models adapted to the Ukrainian language will allow achieving a pilot F1 score of 80-90% for text data and 70-80% for speech data in tasks of classifying destructive content and determining authorship. Product Description popup.authors Ihor R. Korolov Sazhok Mykola Mykolayovych Oksana M. Zuban Valentyna V. Robeiko Yuliia O. Tsyhvintseva popup.nrat_date 2026-02-27 Close