Аннотация:Natural language processing technologies have made great progress today, and considerable merit in this belongs to machine learning, which is used, particularly, for understanding texts. Neural network technologies can be used in any task where text classification is necessary, whether it is spam filtering, fraud defining or credit scoring, determining the mood of a text, or even the author's tendency to be depressed, etc. In almost every paper in the collections of leading linguistic conferences, neural network methods are mentioned. Their popularity is largely due to their ability to find complex, sometimes hidden relationships in the data. However, in order for neural networks to fully demonstrate their practical effectiveness, large amounts of textual data are needed for training. This article tells about the language models used before the neural network revolution, whether it is possible to transfer the text to the computer's memory without losing its structure and semantics, and how a smartphone tells us words in messages, as well as about the use of neural network technologies in linguistics.