Дата публикации: 2021

Дата публикации в реестре: 2022-10-06T22:21:29Z

Аннотация:

Описана исследовательская программа когнитивно-информационного моделирования социальной реальности в массовой коммуникации. Моделирование рассматривается как метод исследования концептуализированных фрагментов социальной реальности, образующих и формирующих медиасферу – иерархическую гиперсеть, образуемую концептами национальной концептосферы и медиасобытиями. В качестве материала используются 207 обучающих текстовых выборок, каждая из которых посвящена одному медиасобытию, входящему в медиаповестку российских массмедиа за годовой период. The article describes the research program of social reality cognitive-information modeling in mass communication. Modeling in this case is considered as a method to investigate conceptualized fragments of social reality, which form and shape the media sphere. The media sphere appears as a hierarchical hyper network, the lower level of which is constituted by the concepts of the national concept sphere, while the upper levels represent media events along with their various clusters. Media events are not understood as mirror images of social reality phenomena, but rather as modes of their existence for mass audiences. In order to select media events for this study, we considered the current media agenda, which consists of issues with a significant number of reposts of relevant messages by a wide range of sources that reflect different views on specific social events. The study included the implementation of the following steps. (1) Media events were collected, and each of them was presented in the form of a sample of unique texts (publications in the media). The material was collected through the web application “Automated Classifier of News Content”, tagging text messages of users of the social network MirTesen. As a result, 207 training text samples were studied, each of which is devoted to a single media event that is included in the media agenda of the Russian mass media over a year. Within the framework of the supervised machine learning methods, each sample of at least one hundred unique publications in the Russian media was used to create a model of the event. Such models were used for automatic tagging classification of new texts placed on the MirTesen platform (automatic classification revealed the relevance of each created media event model). The total amount of training samples was more than 40,000 media texts; the total number of tagged texts is more than 1 million. (2) Each media event was designated by ten most significant concepts represented in the attributes of its model. The model was built on the basis of mathematical processing of a texts sample. The units (words and word combinations), identified as a result of applying the TF-IDF method and ranked by criterion х2, were regarded as concepts due to the fact that in their system of presenting media events they serve as a knowledge construct that sets the perception and understanding of a given event in the interests of leading actors. (3) The concept hyper-network, representing the national sociopolitical concept sphere, which was formed in the Russian mass media in the studied time period, was reconstructed. In total, there are 1,317 nodes (concepts) and 8,822 edges in the constructed hypernet. (4) Clustering using the hypergraph modularity method and subgraph selection was made. (5) The most significant classes of events were described/interpreted. As a result of the research, a conceptual apparatus has been developed for modeling the media sphere, which combines methods of machine learning, linguistic analysis, network analysis and visual analytics. In this work, the classes of media events were reconstructed, their hierarchy (conceptual priorities) was established, and the relationship between media events was revealed. The most significant classes of media events were the EVENTS (10.48% of the vertices of the hypergraph network), PERSONIFIED POWER (9.79%), UKRAINE (8.5%) and COURT (7.2%). The presented research program can be used to solve client-oriented tasks, starting with monitoring the activity of individual actors and reconstructing their conceptual environment in time dynamics, and ending with monitoring and evaluating the state of the media sphere.

Тип: статьи в журналах

Источник: Вестник Томского государственного университета. Филология. 2021. № 72. С. 5-26