FIELD: computer equipment.
SUBSTANCE: group of inventions relates to computing systems and methods for processing natural language. To achieve a technical result, a method is proposed which includes the creation, using a computer system, of a set of feature vectors, learning, using a set of feature vectors, of an autoencoder represented by an artificial neural network, creation, using the autoencoder, the result of the hidden layer and learning, using a learning sample of data, a text classifier. System is also proposed that includes a memory and a processor connected to a storage device. And a permanent machine-readable data carrier is proposed, including executable command, which, when executed by their computing device, lead to the execution of operations by the computing device.
EFFECT: achieving high accuracy of classification when learning a classifier on learning samples of relatively small volume by using the result of a hidden layer of an autoencoder to further learning the classifier.
20 cl, 9 dwg
