The paper presents a conceptual model of a pragmatic-moral discourse as a basis for assembling a training dataset, as well as the results of an experiment of using such data by the Recurrent Neural Network (RNN) to assess how accurately it can determine the attitude of Internet discourse participants towards the pension

Functional features are investigated and shortcomings in the existing process of sending messages about city problems on the portal «Our St. Petersburg» are revealed. The approach to the development of automatic classification of citizens ' messages by existing on the portal categories is described. Based on the reports submitted by citizens in the amount of 1.5 million, training and test samples were formed in the ratio of 80% and 20% of the main volume of texts, respectively. Based on the training data sample and 194 categories, the algorithm of automatic classification was trained using such classical methods of machine learning as naive Bayes classifier, decision trees and artificial neural networks. Using the method of determining the effectiveness of the classification and the test sample, the trained algorithm was tested and checked. The analysis revealed that the algorithm based on the use of artificial neural networks shows the best result among the other methods used. The average classification accuracy of the algorithm was 82%. The trained algorithm was used in the development of an intelligent classifier, which is a web application and implements API mechanisms for interaction with the main modules of the portal information system.

No abstract

No abstract

Классификатор объектов городского хозяйства для данных из социальных сетей Рекомендуемая форма библиографической ссылкиБеген П.Н., Низомутдинов Б.А., Тропников А.С. Классификатор объектов городского хозяйства для данных из социальных сетей // Научный сервис в сети Интернет: труды XXII