Social Network Posts Research Articles

Digital toolset for the response of the public authority institutions to destructive and deviant newsworthy occurrences

The subject of this research is the institutional, organizational and technical peculiarities of the functionality of the system of incident management as a digital toolset for the response of the Russian public authority institutions to destructive and deviant newsworthy occurrences in social networks, instant messengers and online platforms aimed at youth as most active digital citizens. The goal of this article lies in characterizing the current digital situation in Russia, and test the work of the incident management system from perspective of timely determination and processing by the institutional agencies of fake messages aimed or online posts that call to action aimed at young audience for the period from 2012 to 2021. The novelty consists in determining the total number of appeals or complains submitted by the citizens regarding the measures or complacence of the authorities, which have been processed by the incident management system of each of the 85 Regional Management Centers for 2021, as well as in using digital toolset of the incident management system applicable to the data array of the social networks, instant messengers and online platforms (messages that call minors to participate in unauthorized rallies) and deviant newsworthy occurrences (fake sociopolitical posts in social networks and instant messengers), and time interval of their processing by the institutional agencies.

Business continuity in the COVID-19 emergency: A framework of actions undertaken by world-leading companies

The COVID-19 emergency has urged companies to operate in new ways to face supply chain interruptions, shifts in customer demand, and risks to workforce health. The organizational ability to respond to critical contingencies is crucial for business leaders in the perspective of continuing business. In our research, we investigate the actions undertaken by 50 world-leading corporations to respond to the pandemic. Applying content analysis to web pages and social network posts, we extract 77 actions related to 13 sub-areas and integrate these into a five-level framework that encompasses operations, customer, workforce, leadership, and community-related responses. We also describe six illustrative company examples of how the emergency can generate opportunities for creating new value. The study advances the scholarly discussion on the impact of emergencies on business continuity and can help leaders define response strategies and actions in the current challenge.

On the issue of perception of political content in social networks: experience in the application of neurophysiological methodologies

The article examines the peculiarities of the young people perception of political content in social networks. The use of an objective neurophysiological oculography method for the study of visual attention has been proposed. The empirical basis of the study was the analysis of 1500 social network posts of political orientation, to which the tagging of the visual part of the message was applied and basic behavioral statistics were collected, including indicators of communicability, attractiveness and virality. This allowed to conduct a primary classification of political content and form sets of stimulating material for conducting neurophysiological research. The research experiment consisted of consecutive displaying collages of various types of political content (memes, photographs of political leaders, photographs of political events and posters) and simultaneously recording the participants eye movement behavior using a web tracker. The obtained quantitative data was used to determine the trajectory of studying the stimulus material, as well as to assess visual attention when perceiving various types of political content. The study of the stimulus material begins with viewing the image in the upper left corner (0°–45°), then moves to the lower left corner (135°–180°), regardless of the type of content. It was discovered that the average duration of fixations for different types of political content differs significantly (H=8.201, df=3, p=0.042). The most attractive are memes and photographs of political leaders, the least attractive – posters. The first fixations are caused, as a rule, by the pattern of viewing visual information. An increase in the duration of perception of a stimulus leads to a redistribution of the duration of fixations to the images most interesting for the participants in the experiment. It was revealed that the average duration of fixations on the content entails its memorability (images), as indicated by the Pearson correlation coefficient (r=0.564, p=0.023). The obtained data on the visual perception of various types of political content allows to adjust and increase the effectiveness of the communication strategy for promoting certain accounts in social networks by actors of political power.

Semantics of life models in the information field of social networks

In the context of the rapid development of the information environment, a person’s life path, identity, and the living space of an individual are not only presented in real life, but also in the virtual world, as well as in augmented reality. The mixing of virtual and real spaces cannot but affect the daily life of young people. This article attempts to describe and structure the elements of life models as fragments of a life scenario, broadcast through communities in social networks. The social network VKontakte has become a meaningful field for analysis as it is the most popular and frequently used online platform in the information space for the purpose of transmitting interesting and relevant information among Russian peers. The starting point of the study was the assumption that the content of social network posts describing the spheres of life can serve as a basis for building elements of life models broadcast on the Internet. The research consisted of several stages: preparatory (questioning young people about their preferred Internet content) and the main one (semantic and content analysis of the content of social networks). The sample of the preparatory stage was 110 respondents, the average age was 22 years (SD=2.8). The main sample consisted of over 200,000 posts from the top 20 social media communities (with an average of 4 million subscribers, with entertainment and non-themed content) over the past 2 years. With the help of special program codes and computer programs, the posts devoted to the most significant spheres of human activity were analyzed. It was found that the most popular marker words denoting areas of life in posts are work, relationships, and family. The data of automated semantic and content analysis showed that most often, through posts in communities, a model of an active and purposeful person is declared. This is an individual who is ready for action, flexible, mobile with respect to work and self-development, but conservative in regard to building family relationships. The results obtained confirm the assumption that the characteristics of the components of life models are presented most vividly in the posts of Internet communities, are actively broadcast and have their own characteristics. The results of the research data can contribute to the study of the main mechanisms of transmission between contemporaries of characteristics of behavior in real life, values and information through Internet content, as well as the work of practicing psychologists when referring to the profile on social networks or preferred client communities.

Identificación de factores de influencia en el precio de las criptomonedas: evidencia para bitcoin y ethereum

El presente estudio tiene como objetivo el desarrollo de nuevos modelos para determinar con alta precisión los factores que explican el precio de las principales criptomonedas. Para ello, se ha utilizado una amplia base de datos de variables relacionadas con el bitcoin y el ethereum, y se han aplicado técnicas de redes neuronales artificiales. Los resultados obtenidos han permitido identificar que los aspectos relacionados con el número de publicaciones en foros, el volumen de transacciones en blockchain y la tasa de hash proporcionan una excelente estrategia para predecir el precio del bitcoin. También, que el volumen de transacciones, el tamaño de los bloques, las comisiones de mineros y los precios del petróleo son los mejores predictores del valor de mercado del ethereum

Sentiment analysis in textual, visual and multimodal inputs using recurrent neural networks

Social networking platforms have witnessed tremendous growth of textual, visual, audio, and mix-mode contents for expressing the views or opinions. Henceforth, Sentiment Analysis (SA) and Emotion Detection (ED) of various social networking posts, blogs, and conversation are very useful and informative for mining the right opinions on different issues, entities, or aspects. The various statistical and probabilistic models based on lexical and machine learning approaches have been employed for these tasks. The emphasis was given to the improvement in the contemporary tools, techniques, models, and approaches, are reflected in majority of the literature. With the recent developments in deep neural networks, various deep learning models are being heavily experimented for the accuracy enhancement in the aforementioned tasks. Recurrent Neural Network (RNN) and its architectural variants such as Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) comprise an important category of deep neural networks, basically adapted for features extraction in the temporal and sequential inputs. Input to SA and related tasks may be visual, textual, audio, or any combination of these, consisting of an inherent sequentially, we critically investigate the role of sequential deep neural networks in sentiment analysis of multimodal data. Specifically, we present an extensive review over the applicability, challenges, issues, and approaches for textual, visual, and multimodal SA using RNN and its architectural variants.

Irony as a Political Demarcation Tool of the New Russian Nationalists

The article discusses how and why the new nationalists, who call for political self-determination of Russians but share some ideological concepts with liberals, use stiob - a form of ironic parody based on overidentification and decontextualisation, resulting in destruction of the authoritative discourse. Their entertaining, or educational-cum-entertaining projects, located in the gray area between politics and counterculture, strive to undermine domineering political discourses (liberal, neo-Soviet, leftist, official patriotic, and old nationalist) and to go beyond the left-right dichotomy. The author concludes that the main function of stiob and other forms of irony for the new nationalists is negative identification. Ambivalence of the language of stiob simultaneously attracts the target audience of nationalists (“those in the know”) and does not prevent solidarizing with any political platform when needed. The article is based on qualitative analysis of narratives produced by nationalist social media influencers, including fiction, essays, talks, lectures, interviews, live broadcasts, posts in blogs, social networks and messengers.

Social Networks and Atopic Dermatitis: Cross-Sectional Descriptive Study

BackgroundSocial networks have become key tools for finding and disseminating medical information. ObjectiveTo describe the characteristics of social network postings on atopic dermatitis, the content that is posted, and the number of followers of pages, groups, or accounts. Materials and methodsWe selected pages, groups, or accounts related to atopic dermatitis on Facebook, Twitter, and LinkedIn and followed them in January and February 2020. For each site, group, or account we recorded country of origin, year created, purpose, presence of links, provision of a contact email, and number of followers. We also analyzed the topics treated in recent content posts on the pages with the largest numbers of followers in each network. ResultsA total of 257 pages, groups, or accounts were included: Facebook, 171; Twitter, 59; and LinkedIn, 27. Facebook had the largest total number of pages and mean (SD) number of followers: 1416.71 (3722.63). Patient support groups accounted for most of the pages (63%), but businesses or product offers had more followers (P < .035). Of the 909 posts of informative content we analyzed, the most frequent topic was “general information about atopic dermatitis” (27.94%). ConclusionOur findings show the importance of clearly defining the roles and limitations of social media platforms for designing future information campaigns and new patient-centered approaches to reaching patients with atopic dermatitis.

Detecting Suspicious Texts Using Machine Learning Techniques

Due to the substantial growth of internet users and its spontaneous access via electronic devices, the amount of electronic contents has been growing enormously in recent years through instant messaging, social networking posts, blogs, online portals and other digital platforms. Unfortunately, the misapplication of technologies has increased with this rapid growth of online content, which leads to the rise in suspicious activities. People misuse the web media to disseminate malicious activity, perform the illegal movement, abuse other people, and publicize suspicious contents on the web. The suspicious contents usually available in the form of text, audio, or video, whereas text contents have been used in most of the cases to perform suspicious activities. Thus, one of the most challenging issues for NLP researchers is to develop a system that can identify suspicious text efficiently from the specific contents. In this paper, a Machine Learning (ML)-based classification model is proposed (hereafter called STD) to classify Bengali text into non-suspicious and suspicious categories based on its original contents. A set of ML classifiers with various features has been used on our developed corpus, consisting of 7000 Bengali text documents where 5600 documents used for training and 1400 documents used for testing. The performance of the proposed system is compared with the human baseline and existing ML techniques. The SGD classifier ‘tf-idf’ with the combination of unigram and bigram features are used to achieve the highest accuracy of 84.57%.

25.K. Workshop: Unleashing the power of large routinely collected data for the benefit of public health

Abstract As for many other scientific domains, public health is nowadays contemplating the opportunities of using but also the challenges of mastering large routinely collected data ('big data') in order to generate knowledge, and to inform and evaluate decisions and policies. In a wide perspective, “public health relevant big data” extends from the “omics” to social network postings. In this workshop we will concentrate on electronic health records and medical claims data (hereafter referred to as eHRs) which are increasingly being used in public health research and practice. Depending of the national and historical contexts, eHRs encompass details pertaining not only to hospital admissions but also to deaths and infectious diseases registrations, prescriptions and contacts with health services. As a result, eHRs datasets are both long (ie, vast number of records) and wide (i.e., vast amount of information per record). Although the usual public health and epidemiological concepts of inference and causation are still relevant for the analysis of big health data and interpretation of their results, a new set of methodological approaches is also necessary to tackle them. These include for instance quasi-experimental/natural experimental analysis, data mining and machine learning. Recent reviews have demonstrated that these data science terms have corresponding concepts in public health research. This workshop will discuss the use of large routinely collected data in public health with a specific focus on eHRs. The objectives are: to understand the specificity of eHRs in the wider domain of big data; to discuss the challenges imposed when using such data (e.g., data heterogeneity, fragmentation, handling, access, privacy), to discuss how the analysis of big data can assist public health researchers, evaluators and policy-makers; to discuss their advantages and limitations; to outline the methodological, legal and ethical challenges that this entails. Following a brief introduction, the workshop will continue with 5 presentations drawing from a variety of national contexts. The spectrum will spread from countries where eHRs and data linkage have been applied for some years (e.g., Finland, Singapore), to others where development are more recent (France) or even being initiated (Brazil). Together, the presentations will address issues related to the development of big data set, access and training of users and implementation in public health research and practice. The remaining time will be dedicated to interaction with the audience and conclusion. Key messages Access and use of big data in the context of public health practice and research is spreading fast. The challenges ahead consist of maintaining and securing vast amount of heterogeneous health and related data, and in building the capacity to analyse them using old and new analytical approaches.

The spectacularization of violence in contemporary US bare-knuckle fighting

This paper presents an application and a further development of Norbert Elias and Eric Dunning’s (2008) Quest for Excitement thesis through the analysis of the recent phenomenon of the Bare Knuckle Fighting Championship (BKFC) in USA. The research method is based on qualitative data analysis of different digital sources concerning bare-knuckle fighting, ranging from newspaper reports, interviews with promoters and fighters, social network posts, videos, etc. The main finding is that BKFC portrays certain features (e.g. no gloves, blood) that constitute a ‘spectacularization of violence’; it apparently increases the perceived danger of the practice even though it does not increase real harm (e.g. concussions, deaths). BKFC employs other strategies emphasizing control, safety and respectability as a counterweight to obtain an adequate tension balance within the US threshold of sensitivity to violence concerning professional combat sports.

THE RANGE OF COMMUNICATION TOPICS OF THE VISUALLY IMPAIRED IN THE INTERNET COMMUNITIES

Purpose of the study: This study discusses questions associated with the analysis of the most relevant communication topics of sight-disabled people in the Internet communities. According to the World Health Organization, there are about 1.3 billion people worldwide who live with a form of vision disorder. For solving a broad range of social problems of the sight-disabled, interest-based virtual social communities are created in the global internet network. By analyzing the content of these communities, one can judge about the most relevant topics of interest for sight-disabled people.&#x0D; Methodology: In this work, search and study of the Internet communities of sight-disabled people have been conducted in the VK (Russia)] social network. Using the Popsters service of analytics of social network posts and pages, the authors have analyzed the content of the said communities and identified the most relevant topics for this category of citizens. In the research, the hypothesis was used about a stable and significant statistical relation existing between the activity of users in the Internet communities associated with a particular topic and the level of significance of the selected topic for them.&#x0D; Main Findings: The development of information and communication technologies promotes socialization and integration of sight-disabled people into the civic society. The Internet communities of sight-disabled people focused on various topics are registered in the VK social network. The state has to more extensively support and develop technologies contributing to socialization and integration of sight-disabled people into the civic society.&#x0D; Applications of this study: The research findings can be of interest for specialists dealing with problems of socialization and social adaptation of people having vision limitations.&#x0D; Novelty/Originality of this study: It has been found that the Internet communities with the user-supplemented content have the highest quantity of users registered. The topics range of posts in these Internet communities is quite diverse. Posts asserting the traditional life principles: compassion, assistance, overcoming difficulties – win the highest users' response.

Cyberbullying Detection Through Machine Learning: Can Technology Help to Prevent Internet Bullying?

Cyberbullying constitutes a threat to adolescents’ psychosocial wellbeing that developed alongside technological progress. Detecting online bullying cases is still an issue because most of victims and bystanders do not timely report cyberbullying episodes to adults. Therefore, automatized technologies may play a critical role in detecting cyberbullying through the use of Machine Learning (ML). ML covers a broad range of techniques that enables systems to quickly access and learn from data, and to make decisions about complex problems. This contribution aims at deepening the role of ML in cyberbullying detection and prevention. Specifically, the following issues are addressed: i. identifying the features most frequently considered to develop ML models predicting cyberbullying; ii. identifying the most used ML algorithms and their evaluation methods; iii. understanding the implication of ML for prevention; iv. highlighting the main theoretical and methodological issues of ML algorithms in predicting cyberbullying. To answer these research questions, a systematic review of literature reviews, from a total of n=186 records from online databanks, has been conducted. Ten literature reviews have been elected to analyze and discuss evidence about ML preventative potential against cyberbullying. Most of the models used content-based features to predict cyberbullying. The majority of these features includes words written in social network posts, whereas Support Vector Machine, Naïve Bayes, and Convolutional Neural Networks are the most used alghorithms. Methodological and technical issues have been critically discussed. ML represents an innovative preventative strategy that may optimize and integrate educational programs for adolescents and be the starting point of the development of technology-based automatized detection strategies. Future research is challenged to develop algorithms capable of detecting cyberbullying from several multimedia sources.

Redes sociales y dermatitis atópica: estudio descriptivo transversal

BackgroundSocial networks have become key tools for finding and disseminating medical information. ObjectiveTo describe the characteristics of social network postings on atopic dermatitis, the content that is posted, and the number of followers of pages, groups, or accounts. Material and methodsWe selected pages, groups, or accounts related to atopic dermatitis on Facebook, Twitter, and Linkedin during January and February of 2020. For each site, group, or account we recorded country of origin, year created, purpose, presence of links, provision of a contact email, and number of followers. We also analyzed the topics treated in recent content posts on the pages with the largest numbers of followers in each network. ResultsA total of 257 pages, groups, or accounts were included: Facebook, 171; Twitter, 59; and LinkedIn, 27. Facebook had the largest total number of pages and mean (SD) number of followers: 1416.71 (3722.63). Patient support groups accounted for most of the pages (63%), but businesses or product offers had more followers (P <.035). Of the 909 posts of informative content we analyzed, the most frequent topic was “general information about atopic dermatitis” (27.94%). ConclusionsOur findings show the importance of clearly defining the roles and limitations of social media platforms for designing future information campaigns and new patient-centered approaches to reaching patients with atopic dermatitis.

Analysis of Public Perception of the Israeli Government's Early Emergency Instructions Regarding COVID-19: Online Survey Study.

BackgroundOn March 11, 2020, the World Health Organization (WHO) officially declared coronavirus disease (COVID-19) to be a pandemic. This posed challenges to many countries, prominent among which is communication with the public to gain their cooperation. Israel faces different challenges from other countries in its management of the COVID-19 crisis because it is in the midst of a deep constitutional crisis.ObjectiveThe objective of this paper was to examine the response of the Israeli public to the government’s emergency instructions regarding the pandemic in terms of correlations between overall risk perception and crisis management; overall risk perception and economic threat perception; crisis management and compliance with behavioral guidelines; and crisis management and economic threat perception. We also made comparisons between crisis management and spokesperson credibility and between crisis management and the credibility of information sources.MethodsThe sample was established using an online survey that enabled rapid and effective distribution of an online questionnaire during the COVID-19 crisis. The self-selection online survey method of nonprobability sampling was used to recruit participants (N=1056) through social network posts asking the general public (aged ≥18 years) to answer the survey.ResultsParticipants aged ≥65 years perceived higher personal risk compared to those aged 18-30 years (mean difference 0.33, 95% CI 0.04-0.61) and those aged 46-64 years (mean difference 0.38, 95% CI 0.12-0.64). Significant correlations were found between overall risk perception and attitudes toward crisis management (r=0.19, P<.001), overall risk perception and economic threat perception (r=0.22, P<.001), attitudes toward crisis management and compliance with behavioral guidelines (r=0.15, P<.001), and attitudes toward crisis management and economic threat perception (r=–0.15, P<.001). Participants who perceived that the prime minister was the most credible spokesperson evaluated the crisis management significantly higher than all other groups. The crisis management was evaluated significantly lower by participants who stated that infectious disease specialists were the most credible spokespersons. Participants for whom the Ministry of Health website was the most credible source of information evaluated the crisis management higher than all other groups. Participants for whom scientific articles were the most credible source of information evaluated the crisis management lower than those who perceived that the WHO/Centers for Disease Control and Prevention websites or Ministry of Health/hospital websites and health care workers were the most credible.ConclusionsThe higher the public trust and evaluation of crisis management, the greater the compliance of the public with guidelines. It was also found that crisis management and information cannot be approached in the same way for the overall public. Furthermore, unlike other epidemics, the COVID-19 crisis has widespread economic and social consequences; therefore, it is impossible to focus only on health risks without communicating economic and social risks as well.

A Framework for Detecting Intentions of Criminal Acts in Social Media: A Case Study on Twitter

Criminals use online social networks for various activities by including communication, planning, and execution of criminal acts. They often employ ciphered posts using slang expressions, which are restricted to specific groups. Although literature shows advances in analysis of posts in natural language messages, such as hate discourses, threats, and more notably in the sentiment analysis; research enabling intention analysis of posts using slang expressions is still underexplored. We propose a framework and construct software prototypes for the selection of social network posts with criminal slang expressions and automatic classification of these posts according to illocutionary classes. The developed framework explores computational ontologies and machine learning (ML) techniques. Our defined Ontology of Criminal Expressions represents crime concepts in a formal and flexible model, and associates them with criminal slang expressions. This ontology is used for selecting suspicious posts and decipher them. In our solution, the criminal intention in written posts is automatically classified relying on learned models from existing posts. This work carries out a case study to evaluate the framework with 8,835,290 tweets. The obtained results show its viability by demonstrating the benefits in deciphering posts and the effectiveness of detecting user’s intention in written criminal posts based on ML.

Exploring Impact of Age and Gender on Sentiment Analysis Using Machine Learning

Sentiment analysis is a rapidly growing field of research due to the explosive growth in digital information. In the modern world of artificial intelligence, sentiment analysis is one of the essential tools to extract emotion information from massive data. Sentiment analysis is applied to a variety of user data from customer reviews to social network posts. To the best of our knowledge, there is less work on sentiment analysis based on the categorization of users by demographics. Demographics play an important role in deciding the marketing strategies for different products. In this study, we explore the impact of age and gender in sentiment analysis, as this can help e-commerce retailers to market their products based on specific demographics. The dataset is created by collecting reviews on books from Facebook users by asking them to answer a questionnaire containing questions about their preferences in books, along with their age groups and gender information. Next, the paper analyzes the segmented data for sentiments based on each age group and gender. Finally, sentiment analysis is done using different Machine Learning (ML) approaches including maximum entropy, support vector machine, convolutional neural network, and long short term memory to study the impact of age and gender on user reviews. Experiments have been conducted to identify new insights into the effect of age and gender for sentiment analysis.

Investigating classification supervised learning approaches for the identification of critical patients’ posts in a healthcare social network

Nowadays, Healthcare Social Networks (HSNs) offer the possibility to enhance patient care and education. However, they also present potential risks for patients due to the possible distribution of poor-quality or wrong information along with their bad interpretation. On one hand doctors and practitioners want to promote the exchange of information among patients about a specific disease, but on the other hand they do not have enough time to read patients’ posts and moderate them when required. In this paper, we investigate and compare different supervised learning classifiers that we adopted for the classification of critical patients’ posts who can trigger the intervention of the medical personnel. In particular, by considering different Bayesian, Linear and Support Vector Machine (SVM) classifiers we analyze their accuracy considering different n-grams datasets preparation approaches in order to identify the best approach for the identification of critical patients’ posts in a Healthcare Social Network.

МОВНА ТОЛЕРАНТНІСТЬ ТА МОВНА ГРА В СУЧАСНІЙ ПОЛІТИЧНІЙ РЕКЛАМІ

The features of linguistic tolerance and the “play on words” in contemporary political advertising are explored in the article. Different interpretations of the tolerance concept are considered. The issue of tolerance-intolerance is updated. The concept of “communicative tolerance” is outlined and its necessity in the field of political advertising is clarified. Correct forms and means of political advertising are highlighted. The own communication space of the party HOLOS (translated as Voice) was identified, the communication of the ES (translated as European Solidarity) political party in the election campaign 2019 was analyzed. On the basis of posts in social networks there are revealed the tolerant forms of appeal, the questions, the deminitivs, that perform an emotional and evaluative function. The role of exclamatory, ninterrogative sentences in political advertising is investigated. The list of complementary statements, statements addressed to the voter and some instruments of self-presentation are highlighted. Two types of complimentary statements are qualified: those that convey a positive overall value and those with a metaphorized value. Complementary statements have been found to include those expressing the sender’s need for the addressee. A number of statements with an emotional tone have been identified. It has been found the usage of appeals with formulas which provoke the voter to act emotionally. The “play on words” as the tools for influencing on the recipient are explored. Emotional-evaluative vocabulary is analyzed, which expresses approval, criticism, sympathy and metaphors, through which the text receives originality. It’s found out that repeated rhetorical questions and appeals do stimulate communication. There are analyzed the lexems which are metaphorized in many combinations and receive new meaning. Political advertising highlights vocabulary, semantics, and phonetic-graphic ways of playing on words. An example of a phonetic-graphical language play is the use of the root “HOLOS” in many lexems. As an example of a lexical-semantic way of conducting a language game is the campaign “to take the Council under control” (“take into the arms”). The “play on words” is an example of constructive communication in political advertising, an alternative to anti-advertising. This is an example of a tolerant dialogue with a voter.

ПРИМЕНЕНИЕ ЯЗЫКОВЫХ МОДЕЛЕЙ WORD2VEC И BERT В ЗАДАЧЕ СЕНТИМЕНТ-АНАЛИЗА ТЕКСТОВЫХ СООБЩЕНИЙ СОЦИАЛЬНЫХ СЕТЕЙ

The paper proposes an original algorithm for the formation of a training sample for a neural network that provides a sentiment analysis of text posts in social networks. A feature of the algorithm is the use of the extended Russian-language semantic thesaurus WordNetAffect and the expert dictionary of author’s symbols for expressing emotions. In addition, the paper describes the application of a neural network based on the LSTM architecture to determine the emotional coloring of text messages on a social network using two text vectorization algorithms “word2vec” and “BERT”. As a result of the experiments, an indicator of the accuracy of determining the emotional coloring of messages of 87% was achieved using lemmatization as a text preprocessing algorithm and the BERT algorithm when converting it into a vector.