The first two months in the war in Ukraine through topic modeling and sentiment analysis

In recent years, social media remained not only an environment for expression, but it also became an active battlefield where its users emulate their thoughts, feelings, beliefs, and experiences about the ongoing conflicts and wars, and where sometimes social manipulation techniques become difficult to identify, tackle, and counter. Currently, proper social media analytics solutions are still incipient regarding the ongoing war in Ukraine, but they are much needed. Consequently, this research aims to extract and analyze the topics discussed and the sentiments experienced by Ukrainian Telegram users on data collected in the first 2 months of war in Ukraine through a design science research in a data science methodological approach carried out in a multidisciplinary stance contributing to the ongoing strategic, socio-ethical, and technical discourses and efforts. En los últimos años, las redes sociales no sólo han continuado siendo un espacio para la expresión, sino que también se han convertido en un frente de batalla abierto donde sus usuarios emiten sus pensamientos, sentimientos, creencias y experiencias sobre los conflictos y guerras en curso, y donde a veces las técnicas de manipulación social resultan difíciles de identificar, atajar y contrarrestar. En la actualidad, las soluciones adecuadas para el análisis de redes sociales son todavía incipientes en relación con la guerra en curso en Ucrania, pero son muy necesarias. En consecuencia, esta investigación tiene como objetivo extraer y analizar los temas discutidos y los sentimientos experimentados por los usuarios ucranianos de Telegram sobre los datos recogidos en los 2 primeros meses de la guerra en Ucrania a través de una investigación de diseño científico en un enfoque metodológico de ciencia de datos llevado a la práctica mediante un planteamiento multidisciplinar que contribuye a los discursos y esfuerzos estratégicos, éticosociales y técnicos en curso. 近年、ソーシャルメディアは表現の場であるだけでなく、ユーザーが今現在起きている紛争や戦争についての考えや、感情、信念、経験などを競い合う激しい戦場となっているが、時には、その社会的操作の技術を特定したり、その問題に取り組んだり、対抗したりすることが困難になる場合もある。現在、ウクライナで起きている戦争に関して、適切なソーシャルメディア分析のソリューションは、まだ初歩的なものであるものの、強く必要とされている。そこで、本稿では、ウクライナの戦争の最初の2か月間に収集されたデータを基に、Telegram(メッセージアプリ)でウクライナ人ユーザーが議論したトピックと経験した感情を抽出し、分析する。これは、現在進行中の戦略的、社会倫理的、技術的な議論や取り組みに貢献する、学際的な立場で実施されるデータサイエンス的な方法のアプローチによるデザインサイエンス研究を通じて実施する。 Motto: ‘Nothing will end war unless the peoples themselves refuse to go to war.’ ( Albert Einstein). On March 16 2022, a video of President Zelensky was placed on a Ukrainian news website, Ukraine 24, in which President Zelensky was telling his soldiers to lay down their arms and surrender in the fight against Russia. However, this video of about 1 minute is a deepfake video, built to spread false claims and came to existence since the news agency was hacked (Atlantic Council, 2022). On this point, the Ukrainian Defense Intelligence warned precisely 2 weeks before, on March 2 2022, about the preparations done in this sense by Russia (Defense Intelligence of Ukraine, 2022). Shortly after, President Zelensky posted a video response on Telegram and other social media channels in which he denied this claim and further acknowledging maintaining his position (Allyn, 2022; Techcrunch, 2022). This incident shows that social media and cyberspace in general has become an active battlefield in the ongoing war in Ukraine. Moreover, the (still) decentralized nature of social media, together with its spread and use according to one's interests, pose different types of challenges to both users as well as national and international authorities regarding the content contained, together with its nature, validity, target audience (e.g., public opinion), and further spread. In this context, disinformation or misinformation techniques are used with the possibility of altering content to engage public opinion in discourses and beliefs that, for example, either delegitimize, demonize, argue for use of force, or provide specific arguments as causes for war in the target country (Ciuriak, 2022). The efforts and fight against such social media manipulation techniques already show important results by means of embedding different artificial intelligence (AI) and governance-based countermeasures through social media platforms, public and private institutions, and organizations. However, these techniques continue to advance, and feelings of insecurity and instability continue to grow, for instance, using advanced AI techniques for building deepfake content (Carmack, 2022) such as generative adversarial networks (GANs) and autoencoders. While these are now, in many cases, easily recognizable as being fake, they continue to advance, become more complex and intelligent, and therefore, less and less recognizable by the human eye. The remainder of this article is structured as follows. The present section tackles the motivation behind this research while depicting the research aim it intends to achieve. Section 1 discusses relevant research for this article, considering and positioning this research in the ongoing discourses of the existing body of knowledge. Section 2 addresses the methodological approach considered in this research that is, design science research through data science. Section 3 introduces the dataset used in this research together with the necessary preprocessing steps, and discusses the design choices together with implementation aspects involved in the design and development phase of the solution proposed. Section 4 presents the results obtained and positions these results in the ongoing discourses and context of the war in Ukraine. Section 5 discusses the findings of this research and describes the challenges, limitations, and future work. In this section related studies are discussed in respect to the socio-technical nature of this research. Accordingly, related research focusing (i) on datasets and technical techniques used for implementation, and (ii) on policy and social aspects are discussed. Appraising their borderless nature and possibility of targeting both physical and cyber targets (Maathuis, 2022), the battlefield options for cyber weapons has significantly increased in recent years through altering, disturbing, damaging, influencing, or manipulating both humans and systems via the social media environment (Chen & Ferrara, 2022). Such weapons have both direct and indirect impacts on people's cognition (NATO, 2021), beliefs, and behaviors, as well as on different (inter)national security perspectives, dimensions, and aspects (Maathuis et al., 2018a). The manipulation strategies used in social media include spreading propaganda, driving division and polarization, expanding the context to other media types, spreading altered or false information, and distracting or diverting attention (Chen et al., 2022). In this context, a great deal of effort is allocated to designing, building, and implementing corresponding disinformation and misinformation detecting and countering strategies and techniques (Choraś et al., 2021; Islam et al., 2020). Fortunately, the other side of social media, which resonates with its initial purpose, is (live) information spreading and access, as well as communication with others (Krylova-Grek, 2022; Meikle, 2016): essential aspects in modern human existence and society, especially in difficult moments and crises of different natures. To mention a few here, the Arab Spring or the so-called ‘Facebook Revolution’ in 2011 (Wolfsfeld et al., 2013), the influential campaigns with religious and fighting purposes of different extremist groups on Twitter and Facebook (Klausen, 2015; Pearson, 2018), the social movements and protests during EuroMaidan in Ukraine in 2013–2014 (Onuch, 2015), the annexation of Crimea in 2014 through Ukrainian discourse analysis on IQBuzz and strategic implications (Brodovskaya et al., 2019; Saluschev, 2014), and the proceeding discourses of the ongoing war in Ukraine (Chen & Ferrara, 2022). While on the first examples a rich body of literature can be found, considering several technical, social, political, or ethical aspects and implications, information the ongoing war in Ukraine is scarce, with knowledge being in the incipient phase of its development. This information is much needed to get a clear and transparent view, not only on the events happening but also on the discourses, topics, and feelings expressed by the people experiencing them, their dearest and nearest, and/or impacted by them. In this regard, initial efforts have been made for collecting, analyzing, and interpreting the results obtained with respect to different human aspects correlated to the ongoing war in Ukraine based on data obtained from different social media platforms. Juric (2022) analyzed Facebook, Twitter, Instagram, and YouTube data to gain insights into migration flows and integration trends of Ukrainian refugees, which can be of use to the integration processes into corresponding societies, further improve the response to the ongoing humanitarian crisis, and support the labor market integration. Aguerri et al. (2022) analyzed the information, timeline, and impact of the content of 90 accounts labeled as Russian affiliated of individual journalists and media outlets on Twitter before and after February 28 using the Autoregressive Integrated Moving Average (ARIMA) model, finding an increase in both activity and impact. Haq et al. (2022) acknowledged the increase in both anti-Ukrainian and anti-Russian social media content, and propose a possible solution on requiring stricter information controls, but this would imply establishing who should have the power to define them, as well as how, and in which circumstances, they should be executed. Moreover, a set of valuable Twitter datasets have been proposed for conducting further social media analytics on different aspects of the war in Ukraine. To mention here is the Twitter streaming dataset (Pohl et al., 2022) that uses Ukraine, Russia, and conflict as filtering terms, and which was collected 1 week before and goes for q week after the start of the war for further research purposes. Along the same line, Chen and Ferrara (2022) have collected 63 million tweets between February 22 and March 8 2022, using filtering terms such as Ukraine, Russia, NATO, Donetsk, and Zelensky for supporting the research community in this direction, by further maintaining and updating to understand the role that social media plays with respect to information sharing, influence campaigns, and mobilization in war times. Haq et al. (2022) also gathered Twitter data from the beginning of the war using filtering keywords such as Russia, Ukraine, Putin, Zelensky, and Kyiv without applying language or geo-filters to support the corresponding research community with data updated to the present moment, that is, in May 2022. Moreover, Shevtsov et al. (2022) released a recent Twitter dataset containing IDs from 57.3 million tweets. From the volume analysis conducted by the authors, one can see that the activity of registered users decreases with the volume of or accounts significantly increased and that the sentiments the that for relevant insights are and the topics in the corresponding discourses, and the sentiments or feelings expressed by the (2022) gathered million tweets from February to February 2022, using filtering terms such as Russia and Ukraine, and the data to only tweets in while to analyze not only the sentiments and their the beginning of the but also in the before its of the results obtained by such as and have been considered as contributing to while such as and have been considered as contributing to et al. (2022) an analysis on and media discourses regarding the ongoing war in Ukraine, the that while outlets with the and humanitarian aspects of this the Russian media on the for the and news media on the and of this The the and of and relevant social media datasets and their corresponding discourses and that can support both the research and with direct and information regarding different aspects of the ongoing incipient research on different social media such as Twitter, Facebook, and while information on other social that are used such as which into a 2022), to the of is is, therefore, the of this research to this knowledge using a approach to other studies in this that is, and into their and in the related studies discussed This research aims to a technical solution for topics and sentiments discussed by Telegram users in the context of the ongoing war in Ukraine. into the nature of this that has a direct purpose, and implications et al., et al., the design science research et al., et al., is used in a data science approach Accordingly, a of AI that is, are a multidisciplinary approach by social media, international and security studies and the research The increase in the use of social media is also in crises and conflicts of different and where different social, and are & 2015; & The ongoing war in Ukraine is a in this While different media and efforts have this and continue to with to different of the ongoing a of or more specific and discourses are with respect to different topics and the related in this research such discourses war of and humanitarian and several types of such as or to and or to while recent research and positioning data obtained related to ongoing war discourses from social media such as Twitter and Facebook on such as which is used by is 2022), but much needed as this to a to other social media platforms, and it by the so-called through its and direct to this research tackles this knowledge to topics and sentiments of users from Telegram discourses since the start of the war for 2 months from February to 2022) by building a technical solution in this that the to be is and the of this research is the to its corresponding research is as are the topics and sentiments expressed by Ukrainian users in the first 2 months of To be to understand the context of this research and the necessary from the an literature was In this studies related to this research by building of keywords such as and which are used for building in the of and The for and Accordingly, the related studies include strategic and policy analysis studies as well as technical for and policy through the same AI as discussed in the section of this The solution that this article and is a technical of a of that extract topics and analyze sentiments of users by implementing a of in Accordingly, this is in the section of this The results obtained in this research are and in the and discourses regarding different the ongoing war in Ukraine, while future and research are considering a socio-technical The findings of this research obtained and their and the research conducted are by means of this article and In this Telegram data was collected since the first of the that is, February to 2022, a of 2 from a Ukrainian is private for The dataset in the Ukrainian for which the and body was in a for further a was to the collected using the which a into of the by the using techniques in the language that is, 2022). In this the from Ukrainian that is, to that is, en was for this is in the to the they further This involved (e.g., as well as and (e.g., for set of these is in the into and was to and provide a for that to its in a (e.g., the was carried out to a to its or which is to as the instance, the and are to their and the is the In this the was as in on the literature conducted considering the of AI techniques in social media and considering their in in other or related as discussed in the section of this the techniques are further and of these techniques is further and the results obtained from their implementation are and discussed in the The is important conducting a a language where different artificial intelligence techniques are for example, or used to different social media aspects and 2022). In this topics are to which and of topics to a the implementation, the with was used & In a this the of the body using a analysis to provide less or to the that have less by use of the 2020). In advance, the of topics that be should be and they are by their This that the more important is a for a specific the its will for insights into the ongoing war discourses, have been topics and topics the (i) topics have been considered and a direct between and events was from the initial and (ii) topics for the have been considered to present a more of topics and to corresponding for this is in The of the is by the This of language where different or techniques are being to and human and regarding different topics, or et al., 2021; et al., in respect to the sentiments regarding the ongoing war in Ukraine, in this these are and as being or both and possible and et al., which further and from which was its as in In the the dataset was and the implementation of that is, and analysis have been discussed in The results obtained for both techniques are and in the context of the ongoing war in Ukraine. the topics discussed by Ukrainian the first to topics to on the of the topics and the to topics for the to a into the important topics discussed in the topics in both have been further and the results obtained to here to for the analysis it was to the into and present the results obtained in the 3 first of the that is February 2022, March 2022, and 2022. In the shows the of with the of and depicting the shows the topics, together with their corresponding for February 2022, and their corresponding together with the of are in 8 shows the topics together with their corresponding for March 2022, and their corresponding on the side together with the of are in shows the topics together with their corresponding for 2022, and their corresponding together with the of are in from the topics for the that is, from February to 2022, topics to the them, together with their corresponding and against the of to show the of topics The topics are in where the of shows the of and the the the is the in From the results the topics and the first February 2022, for example, Zelensky, and for March they for example, NATO, and while for they for example, and In this several aspects are that the general is which can be by the impact that it on Ukrainian and on users on social focusing on the first of the a direct between the topics and the general feelings of the Ukrainian and the initial and discourses from international and media as well as and from outlets and on aspects such as further of Russian and through in the and the and the of the power Council, 2022; Ukraine, 2022; NATO, 2022; 2022; 2022; 2022). are also with studies social of this war for instance, and Accordingly, (2022) that the ongoing war in Ukraine a for and in which in was and the between the and Russia are In the (2022) the in the and strategic between the on and such as the one in and the on a in this the on and 2022). Such not only have a direct impact on the and between Russia and but also to a set of general and against Russia for the annexation of Crimea 2022). the same the on Russian such as and is and et al., 2022). these are the physical and experienced by of and their as well as the the Ukrainian et al., 2022). to the data collected in this that is, 2022, which 2 months after the war the topics and include and these of on aspects such as further in for example, and and the in the 2022; 1 2022; President Ukraine, 2022; 2022). To to the relevant studies discussed in to the first and of an important that gathered media and humanitarian attention was the Russian of the in which the in the between the the of purpose, and the of 2022). and after February 24, several on Ukrainian for example, on communication systems of the Kyiv on on for to and and for of Ukrainian and 2022; 2022). show and with targeting Ukraine, such as from and from et al., through the same for the one can see such as Zelensky, and a of that of the and of Ukrainian people and are also in the from the analysis conducted and the results discussed by related studies on this this and the results obtained focusing on the Telegram are to (i) the existing studies on discourse and analysis conducted on other social media such as Twitter, Facebook, and and (ii) further position in the ongoing social, and discourses and The ongoing war in Ukraine its complex by with means and in different and both and targets This targeting human and its corresponding processes such as and is a conflict that the physical of Ukraine through its strategic, and implications, and this from a conflict into an indirect international The ongoing be as the of the a recent war that and other means to strategic conducted in the years through a of means and in Ukraine and as well as through the building and of cyber and information through such as social media and media channels and outlets in general 2015; & 2015; et al., 2022; & 2019; 2022). While countermeasures for social media manipulation from direct in the development of social media platforms, as well as and based on both data and human the is still since are in the and of both and essential discourses and topics together with their corresponding sentiments would an intelligent, and transparent approach for and/or and and/or countering social media manipulation in of several studies have in this on these in to the ongoing war in Ukraine and events carried out in the on social media such as Twitter, Facebook, and while a studies on other such as these as a for the implications, and motivation behind this in this the aim is to a technical solution for topics and sentiments of Telegram users in the context of the ongoing war in Ukraine. To this the design science research et al., et al., is used in a data science approach for building corresponding a multidisciplinary Accordingly, data between February and 2022, is collected and for building and analysis for the content as well as for the The results obtained aspects the war in Ukraine, being related to and and as topics a the that to the and of the that is, and a with and media corresponding discourses and results be further considered in research for (i) and discourse analysis for general topics and sentiments and of (ii) social, and impacts based on research on sentiments and of users about aspects such as and both and the topics contributing to different types of sentiments and of the same these results be further considered in research for (i) insights into discourses and sentiments for communication as well as for public security campaigns against social manipulation and (ii) the involved in such a complex and conflict such as the one ongoing in Ukraine. the and not the of a (