Контент-аналіз соціальних медіа методами машинного навчання

Abstract
Робота присвячена проблемам проведення контент-аналізу електронних соціальних медіа. Проаналізовано особливості соціальних медіа та їх відмінності від традиційних засобів масової інформації. Визначено основні завдання, що дозволяє вирішувати застосування контент-аналізу соціальних медіа для бізнесу. Обґрунтовано використання технології машинного та глибокого навчання для створення систем контент-аналізу. Запропоновано застосування технології Transfer Learning для перенесення знань із переднавчанених мовних моделей на інший домен або іншу мову, зокрема, українську з інших слов’янських мов. Introduction. During the last two decades, the rapid development of social media has caused a revolution in means of communication in modern society. Therefore, a significant number of the world's leading companies began to rebuild their business models using the capabilities of modern means of communication through social networks and other platforms, for which content analysis technologies are successfully used. Purpose. The purpose of the article is to develop methodological principles for conducting content analysis of electronic resources (social media) based on using Artificial Intelligence technologies, in particular, Machine and Deep Learning. Results. The paper analyzes the phenomenon of social media and identifies the key factors that determine the effectiveness of their use for both business and consumers. Based on this, the paper explores the features of the content analysis of social media, which take into account their mass character, as well as the presence of large arrays of unstructured information. The large amount of information on various electronic platforms requires adequate means for their monitoring and processing, analysis of content as well. To solve these problems, the paper substantiates the use of modern Natural Language Processing technologies based on Machine and Deep Learning approaches. An alternative to the existing services for content analysis is developing systems based on such tight forwarding motion models, like BERT (provided by Google), or GPT-3 (provided by Open AI), which was implemented on the Transformer Deep Neural Networks architecture. The article also proposes the use of Transfer Learning technology to transfer knowledge from pre-learned language models to another domain or another language, in particular, Ukrainian from other Slavic languages. Originality. The main findings of this paper are the follows: (i) the main advantages of using social media for businesses and consumers are substantiated; (ii) the characteristic features of conducting content analysis in social media are determined; (iii) the advantages and disadvantages of using Natural Language Processing methods for solving problems of content analysis in social media are shown; (iv) Transfer Learning approach to transfer knowledge from pre-learned language models to another domain or another language, in particular, Ukrainian from other Slavic languages has been proposed for solving content analysis tasks. Conclusion. The accumulation of a sufficient amount of training data, the development of multi-core CPU and graphics processors, as well as the formation of powerful pre-trained language models and the development of effective algorithms for processing extremely large amounts of information are factors that determine the efficiency of the use of Machine and Deep Learning technology for content analysis tasks in recent years. Therefore, the development of computer systems for content analysis of social media, in particular, using modern technologies of Artificial Intelligence (Machine and Deep Learning), does not lose its relevance and requires further research.
Description
Keywords
контент-аналіз, соціальні медіа, машинне навчання, глибоке навчання, content analysis, social media, Machine Learning, Deep Learning
Citation
Ахмедов Р. Р. Контент-аналіз соціальних медіа методами машинного навчання / Ахмедов Ренат Рамазанович, Дербенцев Василь Джоржович, Віталій Сергійович Безкоровайний // Вісник Черкаського університету імені Богдана Хмельницького. Серія: Економічні науки : зб. наук. пр. / М-во освіти і науки України, Черкас. нац. ун-т ім. Богдана Хмельницького ; [редкол.: Є. М. Кирилюк (голов. ред.) та ін.]. – Черкаси : Черкас. нац. ун-т ім. Б. Хмельницького, 2022. – Т. 26, № 3–4. – С. 48–57.