Tokenization of Political Texts in BERT Models Using ICF+ Ontologies

The design of machine learning language models, as well as their ensembles, used in complex analytics of news texts of domestic and Western electronic media is considered. An example of software implementation of a new language neural network model with problem-oriented ontological tokenization is given. The language used as tools is Python v.3.10, Anaconda v.2.1. The effectiveness of the approach in comparison with the best foreign analogues is confirmed by a series of experiments using the example of classifying news articles according to their ideological orientation into Western and English-language Russian ones.