DEVELOPMENT OF SOFTWARE COMPLEXES OF TEXT DATA RECOGNITION

The technique of development of software complexes of recognition based on neural networks. It is shown that the training eliminates the need to select key features, their importance and relationships between characteristics. But, nevertheless, the original representation of the input data greatly affects the quality of solutions. The neural network has good generalizing ability, i.e. can successfully disseminate the experience gained on the target training set, the whole set of images