Neural network technology for detecting errors in text documents

Abstract

The goal of this work is to develop a modified technology for error detection in text documents using a multilayer perceptron neural network. The proposed technology is implemented using an autoassociative neural network trained on a corpus of parallel texts containing distorted and corrected sentences. The feasibility of using the principal component method for preprocessing input text data is investigated. The test results confirm the effectiveness of the application of the studied technology for detecting errors in polythematic text documents.