Риск-анализ контента социальных сетей на основе нейросетевой классификации эмоциональной окраски текста сообщений
Работая с сайтом, я даю свое согласие на использование файлов cookie. Это необходимо для нормального функционирования сайта, показа целевой рекламы и анализа трафика. Статистика использования сайта обрабатывается системой Яндекс.Метрика
Научный журнал Моделирование, оптимизация и информационные технологииThe scientific journal Modeling, Optimization and Information Technology
Online media
issn 2310-6018

Risk analysis of social media content based on neural network classification of a message text emotional coloring

idRazinkin K.A., Sokolova E.S.,  Savishchenko D.N.,  Chapurin E.Y. 

UDC 004.056; 004.032.26
DOI: 10.26102/2310-6018/2021.35.4.034

  • Abstract
  • List of references
  • About authors

One of the promising areas of Data Science within the framework of practice-oriented approaches to the analysis of social networks (Social network analysis) from the point of view of network users’ (agents’) opinion formalization is a class of content analysis methods designed for automated identification of emotionally colored vocabulary in texts and emotional evaluation of authors in relation to the objects referred to in the text. With the help of such an analysis, it is possible to study an array of messages and other data and determine how they are emotionally colored - positively, negatively or neutrally. The article offers a comparative analysis of two approaches to the study of text sequences classification possibilities depending on their emotional coloring: one by means of a recurrent neural network (RNN) and another involving graph convolutional networks (GCN). The first approach is implemented through deep learning utilizing the Deep Learning Designer tool (MathWorks © MATLAB R2021b). The second approach is based on the application of convolutional graph neural networks for text classification. GCN implementation is carried out in Python using the appropriate set of libraries for data analysis. In addition, the paper shows that the resulting model can be used in risk assessment, where the resulting value serves as a correction factor in calculating the risk of user involvement. Based on the results of the two approaches comparison, it is shown that when using GCN, the percentage of training data decreases, which indicates the sensitivity of the method to a smaller amount of training data, while the accuracy of the model increases with comparable configurable training parameters

Keywords: emotional coloring of the text, recurrent neural network, deep learning, graph convolutional networks, risk analysis

For citation: Razinkin K.A., Sokolova E.S., Savishchenko D.N., Chapurin E.Y. Risk analysis of social media content based on neural network classification of a message text emotional coloring. Modeling, Optimization and Information Technology. 2021;9(4). URL: https://moitvivt.ru/ru/journal/pdf?id=1105 DOI: 10.26102/2310-6018/2021.35.4.034 (In Russ).

854

Full text in PDF

Received 05.12.2021

Revised 25.12.2021

Accepted 30.12.2021

Published 31.12.2021