Authors

The article focuses on the study of the modus category of evidentiality in media discourse using the method of language modeling. Particular attention is given to the development of a neural network model capable of predicting the progression of evidential statements (ES) in accordance with the communicative intentions of the author or speaker. The application of the linguosynergetic approach enabled the differentiation between basic and extended evidential semantics, as well as the description of the interrelation between evidentiality and other modal categories, such as authorization, approximation, perception, modality, negation, evaluativity, personal reference, persuasiveness, expressivity, emotivity, and temporality. The research is based on a corpus of English-language media discourse (BNC), annotated according to the phasal structure of EUs. The typical combinations of modal categories identified through combinatorial analysis served as the foundation for training the language model, which is capable of predicting modal development in discourse context and effectively recognizing both explicit and implicit cases of evidentiality. The results confirm the effectiveness of language modeling in tasks related to the recognition, description, and interpretation of polymodal semantics and highlight the potential of neural network methods for the linguistic analysis of complex discursive structures. The proposed methodology can be applied in the development of automated linguistic expertise systems for media texts, including the detection of manipulative strategies.