Text classification using convolutional neural network committee training

The method of classification of textual information based on the apparatus of convolutional neural networks is considered. The word-by-word text conversion into dense vectors is considered. Testing was conducted on the text data of the sample “The 20 Newsgroups”, this sample contains texts distributed in 20 classes. The accuracy, the best of the convolutional neural network used in this work, on the test sample was ~ 74%. The accuracy of voting of neural networks using the Bagging algorithm was ~ 81.5%. Based on the review of similar solutions, a comparison was made with the following text classification algorithms: using the support vector machine (SVM, 82.84%), naive bayes classifier (81%), k nearest neighbor algorithm (75.93%), a bag of words