UDC 007:681.512.2
NEURAL NETWORKS OF NEW MULTIPOLAR WORLD: CLASSIFICATION OF ELECTRONIC NEWS
I. Yu. Kashirin, Dr. Sc. (Tech.), full professor, RSREU, Ryazan, Russia;
orcid.org/0000-0003-1694-7410, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
A new technology for automatic identification of news media materials is being considered, dividing them into pro-Western content and the articles from independent states. Domestic publishers are used as content sources: RT, Meduza, Kremlin, Globalaffairs, Themoscowtimes, RussiaBeyond, Rossiyasegodnya, Interfax, SputnikInternational and others. Western information publications are represented in the study by publishers: msnbc, bloomberg, cnn, springer, nbcnew, thrguardian, Facebook, nytimes, france24 and others. The theoretical foundations used in the study are based on the concept of neural networks with concentration of attention, namely a bidirectional model with Bert transformer architecture. The experimental part of the material uses software oriented to Python v.3 (Anaconda 3). The software implementation of text corpus collection, subsequent content processing and neural network analysis involves the use of nltk, transformers 4.34.1, BeautifulSoup, wordcloud, BertForSequenceClassification, torch 3.12.4, newspaper, json, tensorflow 2.14.0, accelerate 0.20.1, sklearn. In addition to the listed software, the research uses the htmlgrabber v.2.0 package developed by the author. The performed series of experiments allows us to qualify the presented technology as an electronic news identification technology, which is not inferior in efficiency to the international analogues available today. The aim of the work is to create a new neural network technology used in the automated identification of information content in natural language to classify electronic news into Western and independent.
Key words: : neural Bert networks, natural language analysis, identification of news content, transformer architecture, text corpus assembly, pre-trained models.