Reengineering support and decision-making system based on Twitter analysis
Students Name: Kubinska Solomiia Volodymyrivna
Qualification Level: magister
Speciality: Systems and Methods of Decision Making
Institute: Institute of Computer Science and Information Technologies
Mode of Study: full
Academic Year: 2022-2023 н.р.
Language of Defence: англійська
Abstract: The influence of social networks on the life of an average person is extremely large today [1]. Based on data provided by the Statista source, the Twitter platform was among the five most popular social networks in Ukraine as of 2021 [2]. A large increase in users and the expansion of the Ukrainian-speaking segment of Twitter occurred in the middle of 2022, because Ukrainians needed a platform that would allow them to spread the truth about the war started by russia in Ukraine, without platform restrictions and subsequent blocking of accounts. Another advantage of Twitter is that it is used by many famous and influential people. Thus, regular users can spread their thoughts among major social media, politicians, businessmen, etc. New users of social networks always face two main problems: integration into the right segment of the platform to avoid consuming disinformation [3] and prediction of the reaction of other community members to their posts. These two problems are especially amplified in times of war, when the propaganda of the aggressor country tries to work to reduce the morale of Ukrainians and the sensitive reaction of users to the information and personal opinions they see online. The study object of the master’s thesis is the social network Twitter. Here are the reasons for choosing: the lack of information on the operation of the algorithms presented by the platform, the lack of reliable filtering methods for malicious accounts blocking creates the need for additional analysis, recommendations and the presentation of relevant and, most importantly, safe information to the consumer. The scope of research lies in improving existing methods of analyzing and supporting user decision-making when working with social networks to minimize the spread of disinformation. The goal of research is to create a system for the analysis of "hot" topics and profiles on the Twitter platform and further provide the user with support for decision-making based on his needs. The subject of research is to review the analytics and environment that Twitter algorithms create for the user in order to avoid spreading disinformation and create an opportunity to predict the reaction of the user’s audience on the tweet that the user. Furthermore, it allows the creation of a system which predicts the audience reaction to a user’s tweet. As a result of the conducted research, a reengineering support and decision-making system based on Twitter analysis was developed. The first module allows the system to support the user’s decision in forming its own "ecological environment" in the Twitter network, by creating a list of popular accounts, hashtags and topics that the user can follow and that are relevant. This result was achieved using the Twitter API, which in turn provides access to trends and major events that are currently being discussed on Twitter. Thus, Ukrainian officials who use this platform, volunteers, foundations and other influencers are recommended for user subscription to stimulate algorithms to propose relevant accounts which will form a reliable environment for the user in the social network. The second module of the system allows users to check the tweets they are about to publish for audience reception. This functionality is especially useful for brands as it will help them predict their customers’ reactions to future launches and products. In this way, they will be able to adjust the tweets before publication and make them more relevant for the perception of customers. To achieve the goals of this component, the sentiment analysis of tweets using natural language processing methods is helpful. The NLP is a branch of computer science that works on giving computers the ability to understand human language in written and spoken versions, and the sentiment analysis is a task that is needed to separate subjective qualities from the text, for example, behavior , sarcasm, emotions, suspicion, bewilderment [4]. Dialogflow [5] from Google was chosen as the tool for such analysis. Within the framework of the study, tweets of real users were analyzed. When comparing the results of Dialogflow and the real users reaction to the selected tweets, it was determined that the trained agent showed itself very good in determining the correct user reaction to the tweet. This work provides a basis for further research in this area. For example, the use of mixed approaches (both lexicons and machine learning) to improve the sentiment of posts in social networks. Analysis of user also profiles and posts to detect propaganda. Key words: Twitter, tweet, decision-making system, natural language processing, sentiments analysis, Twitter API, Dialogflow, agent. References: 1. Ngai E. W. T., Tao S. S. C., Moon K. K. L. Social media research: Theories, constructs, and conceptual frameworks. International Journal of Information Management. 2015. Т. 35, № 1. С. 33–44. URL: https://doi.org/10.1016/j.ijinfomgt.2014.09.004 (дата звернення: 15.11.2022). 2. Most popular social media by age Ukraine 2021 | Statista. Statista. URL: https://www.statista.com/statistics/1256255/most-popular-social-media-by-age-ukraine/ (дата звернення: 15.11.2022). 3. Identifying Fake News on Social Networks Based on Natural Language Processing: Trends and Challenges / N. R. de Oliveira та ін. Information. 2021. Т. 12, № 1. С. 38. URL: https://doi.org/10.3390/info12010038 (дата звернення: 15.11.2022). 4. Patil R., Kumar M. SENTIMENT ANALYSIS ON TWITTER POSTS. International Journal of Scientific Research in Engineering and Management (IJSREM). 2022. Т. 6, № 6. doi: 10.55041/IJSREM15023. 5. Sabharwal N., Agrawal A. Cognitive Virtual Assistants Using Google Dialogflow. Berkeley, CA : Apress, 2020. URL: https://doi.org/10.1007/978-1-4842-5741-8 (дата звернення: 15.11.2022).