Development of a trainable neural network for analysis of linguistic data

The description of the developed linguistic application for the recognition of ancient Egyptian hieroglyphs is given. The main methods for creating and training a neural network are considered and a method suitable for the successful operation of the algorithm for recognizing ancient Egyptian hieroglyphs is determined. Based on the analysis of tools for solving the problem, the method of learning with a teacher and the convolutional type of neural network were chosen as optimal for image recognition with the ReLu activation function. In the future, the proposed neural network will find application in the development of a dictionary with a character recognition function.