Hello. The task is to develop a classifier of text according to the style of speech (artistic, publicistic, etc.). It is necessary to do this on the basis of a neural network. I can’t define the characteristics that determine the style of the text and figure out how to translate the source text into a vector of characteristics that the classifier will eventually work with.

How can this be done?

  • As far as I understand, this is precisely the task of the neural network, the algorithm will be developed in the process of its learning. Have you worked with neural networks before? - D-side
  • @Napkatyk No, I very superficially understand the principles of the structure of neural networks. What resource can you recommend to use to put everything in its place? - loginovsky

1 answer 1

Here is the table: text styles (The picture of the table is taken from here )

As can be seen from it, in the texts of different styles, either terms, or descriptions, or references, or slang, or other classes of words, can prevail. Classes of words will determine the neural network after training. It will determine the styles, recognizing the predominance of a certain class or combination of classes of words in the text provided for recognition. The structure of the text will also be among the signs (at least at the level structured / unstructured). And further - literature about neural network training with reinforcement and training with a teacher.

The most important thing is the Error Propagation Algorithm . ( https://habrahabr.ru/post/154369/ - on Habré).