Content analysis methods for estimating the dynamics of facebook groups Rasa Kasperienė Faculty of humanities Tomas Krilavičius Vytautas Magnus univercity Faculty of informatics Kaunas, Lithuania Vytautas Magnus univercity rasa.kasperiene@gmail.com Kaunas, Lithuania tomas.krilavičius@vdu.lt Abstract— The relationship between the content that is Social networks can become a tool for manipulating the generated by the users of social networks and their dynamics masses and fighting wars with little to no cost. has been analyzed by many scholars. However, due to favorable The present paper proposes a framework for carrying out data policies, the majority studies have been carried out by research on posts from Facebook (further FB) groups as a analyzing Twitter data. In addition, such research on Facebook (FB) groups (esp. political) is usually qualitative. The present means to reveal information dissemination and group study analyses the dynamics as well as topic dynamics of radical behavior patterns in communication by information right political groups on FB by employing a quantitative transmission dynamics in groups. In particular, the aim of this research methodology. The current paper draws on a large data study is to analyze the establishment of radical right FB set that is comprised of posts from FB groups. Overall, there are groups in relation to the political events of the time as well as 79 728 posts which are made up of more than 2 million words the dynamics of the most prominent themes by using the data and were generated within the timespan ranging from 2010 to retrieved from FB groups and R toolset. This article 2018. The experimental set up compares the general dynamics investigates the launch of Lithuanian radical right FB groups and the dynamics of activity on four topics in two radical right in a wider political context. It is important to understand the FB groups (i.e., pro-Russian and other radical right) in Lithuania. The results show that the year 2014 was important dynamics and the reasons behind the activity of such groups. for the radical right FB groups in Lithuania. Newly created pro- Another important issue is to pinpoint when the topics Russian FB groups started growing rapidly, whereas the posting discussed in the aforementioned FB groups become relevant activity in other radical right FB groups started to decrease. The and no longer relevant. Finding the answers to these questions topic word Lithuania is relevant for the whole activity time can provide a deeper insight into the social processes of when it comes to all the radical right FB groups. Such topic radical right groups on FB. words as Russia and land correlate with national and Such social networks as FB and Twitter have become the international political crisis. most popular social networks in the world. In 2017, Twitter had more than 330 million active users, whereas FB had more Keywords—Facebook groups, radical right, groups dynamics, timestamp. than 2.13 billion monthly active users with a 14 per cent increase every year [5]. This giant flow of information has I. INTRODUCTION already shown to be useful for event detection [6], identifying In recent years, the European Union has been witnessing public health issues [7], behavioral information propagation the growth of radical political communities throughout [1], community discovery [8], sentiment analysis [9], Europe, including Lithuania. Many European countries are identification of communication roles [9], and recently as a witnessing elections in which people vote for far-right and means to aid political uprising [10] as well as a medium that nationalist parties, even though they are at the opposite ends can help to pinpoint and analyze the act of triggering an of a wide political spectrum. The migrant crisis accelerated a (upcoming) uprising [11]. backlash against the recent political balance, but the wave of II. DATA SET discontent also taps into long-standing fears about globalization and dilution of national identity. The increase FB groups are the place for small group communication in the percentage of radical wing voters substantially and for people to share their common interests and express surpasses the percentage of immigration inflow [1]. their opinion. Such groups allow people to come together The political radicals are more avid and enthusiastic to around a common cause, issue or activity in order to mobilize, adopt new technology and have thus found the virtual space express their objectives, discuss issues, post photos and share to be a uniquely useful place [2]. Through membership in related content [12]. All FB groups have a title and a group groups, one can define and confirm his/her values and beliefs description that indicate the common cause of group activity. through incoming information or discussion. When members FB groups can be public or closed. In the first scenario, every of such groups face uncertain situations, they can gain FB user can access group content. In the latter, content can be reassuring information about their problems and find security accessed only with a permission given by the group in companionship [3]. It is also important to highlight the fact administrator. To comply with the ethical aspects of doing that social media provides fertile ground for the research, the present study only reports on data that has been dissemination of propaganda and disinformation as well as retrieved from public FB groups. the manipulation of people’s perceptions and beliefs [4]. The data were downloaded by using the FB graph API [13]. The Graph API is created to get data into and out of the FB platform. This FB platform uses low-level HTTP-based © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0) API access that can only be obtained by a user who is 74 registered as a FB developer. For API requests, it is necessary TABLE I. to have the access token (app id) together with its app Short data info password and the access token. Posts published period 4th of March 2010 – 1st of January FB API requests return the following group data [13]: 2018 Number of posts 79728 post author id (from_id) as numeric string, post author name Download date 12th of February 2018 (from_name) as string, post text (message) as string, post Max length of word 15 symbols creation date (created_time) as string, post type (type) as Min length of word 1 symbol string, link in the post (link) as string, post id (id) as numeric Lithuanian is a highly inflectional language, i.e. there string, daily entry (story) as bool, likes count (likes_count) as are two grammatical genders for nouns and there are three number, comment counts (comments) as number, shares genders for adjectives, numerals, participles, and pronouns. count (shares_count) as number. Every word must follow the gender and the number of the To analyze the posts of FB groups as a means of noun. All these features produce a substantial number of information dissemination together with the patterns of group inflective forms of lemma. To avoid any loss of data, the behavior in terms of communication by information lemmatization of the texts in FB posts was not used. transmission dynamics in groups, the following subset of data was used: III. METHODS post text (message) as string, post creation date To analyze the dynamics of the topics discussed in (created_time) as string, post id (id) as numeric string. groups, the most frequent words were employed as features To handle the large dataset more efficiently, fingerprint of [16]–[18]. In addition, social networks post timestamp each FB group was created, and it contains the names and ids modelling was applied to analyze the behavior of online users as well as the names of the dataset that come from the FB [19], [20]. This paper proposes to study the posts from FB groups in focus. groups as a means of information dissemination and group The radical right groups on social network FB were behavior patterns in communication by information identified through the Facebook search engine. The transmission dynamics in groups. The proposed approach is supporters of radical right diverge from other individuals based on the following observation: the amount of though manifestation of nationalism, strong nation [14] and information passed from one period to another in the social xenophobic ideology [15]. Nationalistic ideology relates to network may be quantified in different ways. For example, in ethnocentrism and Euroscepticism. Xenophobic ideology the dataset of FB groups, the amount of information can be relates to anti-immigration policy, hostility to ethnical quantified by the time that passed from one post’s appearance minorities, and intolerance to sexual minorities. To identify to other. The quantity of published group posts in a social radical FB groups by using the FB search engine, their most network by looking at the time frame can show group prominent characteristics were taken into consideration, and behavior. based on that, the following keyword list was compiled: To grasp the information transmission when it comes to Lithuania, Lithuanians, land sale, European Union, NATO, the group dynamics, the datasets of FB groups were expanded refugees, refugee crisis, Muslims, Jewish restitution, Jew, by adding fingerprints entries. Let a pro-Russian FB groups Russian, Roma tabor, gay pride, gay mountaineering. More dataset be denoted by D1 and another radical right group than 20 most recent posts in each group that match the dataset be denoted by D2. W represents time window (W = 6 keywords were analyzed. After the analysis that aimed to months). Denote each Facebook post as eij, where i =1 pinpoint the FB groups which openly exhibit radical represents that a post belongs to D1 and i=2 represents that ideology, only 10 groups that proved to endorse radical post belongs to D2; 𝑗 = 1; 𝑛𝑖 where ni is the number of posts ideology were chosen for a more in-depth analysis. The FB in group Di. Each post eij consists of pij, tij, gij. Each post pij, group selection criteria were the following: the presence of consists of a set of words pij = (wij1, wij2 … wijk), where k is radical left ideological features on group titles, description the number of words in pij. and latest posts, the size of the group (more than 100 members), activity – the most recent post published at least 2 days prior to the analysis. The data retrieved from FB groups were divided into two datasets, pro-Russian and other radical right groups. The analysis reveals that some radical right groups in addition to the nationalistic ideology manifest pro-Russian and pro- socialism ideology. Even though in some cases the titles and descriptions of the group’s manifest nationalism and the idea of strong Lithuania, there was also support for Russian politics or a sense of nostalgia for the Soviet Union. Each data set is comprised of five FB groups. As was previously indicated, to be able to handle such large amounts of data, the datasets were supplemented with additional records, i.e., the group and cluster ids. The dataset of pro- Russian FB groups consists of more than 70 150 posts. The Fig. 1. Datasets of Facebook groups with expanded fingerprints entries second dataset, i.e. that of the other radical right groups, is comprised of 9 578 posts. The former dataset of groups has To compare the dynamics of the users in the two 13 940 members, whereas the latter has 6 126. datasets, the transformed dates were stored from string to 75 POSIXct objects. To transform the dates, Lubridate [21] pro-Russian FB groups is 44 per cent greater than it was package for R was used. In order to visualize the distribution initially in 2014. of groups’ activities through time, ggplot2 [22]] package for R was used. It helps to visualize the distribution of a single TABLE II. continuous variable by dividing the x-axis into bins and The dynamics of radical right Facebook groups activity counting the number of observations in each bin. To make the Radic Year text of the post tidy and the datasets lighter, Tidytext [23] and al right 201 201 201 201 Stringr [24] packages for R were employed. By using these group 0 1 2 3 2014 2015 2016 2017 packages, English and Lithuanian stop words were removed. s To estimate the dynamics of the topics in the collected posts, Pro- 1044 1703 1925 2341 each entry (in form of sentences) was split into words. Once Russia 0 0 0 0 7 3 7 3 n again, to keep track of data, every split word was Other 116 supplemented with a post and dataset id, group name, and 58 311 604 2997 791 1914 1739 4 timestamp entries. Total 58 311 604 116 1344 1782 2117 2515 4 4 4 1 2 IV. EXPERIMENT During the course of the Ukrainian crisis, the role of The preliminary analysis identified two types of radical actual military interventions has remained low in comparison right ideology in FB groups under investigation. The to different tools of asymmetric warfare (e.g., information visualization in “Fig. 2” compares the dynamics of pro- warfare, economic measures, cyber war, and psychological Russian and other radical right groups’ activity. It includes war on all levels), often referred to as hybrid warfare [25]. the posts (message) of both groups’ members and post This cyber war passed national or post-Soviet Union borders creation time (created_time). It also shows the peak activity more widely and the spread of fake news reached the western periods that can be noticed in the datasets (within a time world. The conflict in the Ukraine re-awakened Russian window of six months). propaganda. For example, Twitter analyst Lawrence Alexander has identified an increase in bot registration coinciding with the start of the Euromaidan protests on 2013/2014 year in Ukraine and subsequent armed uprisings by pro-Russian militants in Eastern Ukraine in early spring of 2014 [26]. Lawrence’s investigation corelates with rise of pro-Russian Facebook groups in 2014. Prior to 2014, on FB there were only radical right groups with low activity, but after 2014, the situation has changed. The activity of the newly created pro-Russian groups started rapidly growing. According to NATO Strategic Communications Centre of Excellence, some techniques, such as Russian propaganda techniques in particular, are used for achieving psychological influence and manipulation on social media [27]. One of such techniques is the mass-generated content which is used in order to spread manipulative messages and minimize alternative voices. To analyze the dynamics of the most relevant topics in the groups, four keywords were chosen, namely, Lithuania, Fig. 2. The dynamics of radical right groups’ activity on Facebook Russia, land, and sky. The words Lithuania, land and Russia were chosen for this experiment based on the previously The experiment shows that the activity of radical right defined most prominent characteristics of radical right FB groups starts in 2010, whereas pro-Russian groups groups. The word Russia also was chosen in order to assess emerge on FB four years later, in 2014. The pro-Russian and compare the dynamics of topics discussed by pro-Russian groups that were created on the same year reached three times and other radical right in relation to the country. The neutral greater activity compared to other radical right groups on FB. word sky was chosen to reveal whether there is any space for From 2014 to 2017, the activity of pro-Russian groups has neutral topics in the datasets of radical right groups. been increasingly growing. The activity has reached the maximum peak in 2017 with 23 413 posts per year “table 2”. TABLE III. Until 2014, the radical right groups were witnessing the The dynamics of the word Lithuania in the posts of radical right growth of posting activity every year, too. The year 2014 was groups on Facebook important for the radical right FB groups as new ideology- Radica Year l right 201 201 201 201 following radical right groups started appearing and rapidly groups 0 1 2 3 2014 2015 2016 2017 growing. After the appearance of pro-Russian groups on FB, Pro- 0 0 0 0 219 517 441 530 the data spread in other radical right groups started Russian 5 3 0 2 decreasing, but the activity of pro-Russian groups on FB Other 51 74 446 657 189 520 740 857 0 increased each year. This is evident because in 2015, the Total 51 74 446 657 408 569 515 615 activity of pro-Russian groups on FB was 61 per cent greater 5 3 0 9 than in previous year. Finally, in 2017, the posting activity in 76 The topic word Lithuania is relevant for all the radical words shows that from the beginning to the end of 2018, the right groups “Fig. 3”. This word in the posts of FB groups words Russia appeared 16 times more than in other radical appears more than 22 300 times throughout the whole period right groups. The word count indicates that the word appeared of groups’ activity “table 3”. In 2014, the topics that 2 864 times in pro-Russian and 178 times in other radical mentioned the word Lithuania were mostly discussed by right groups “table 4”. newly created pro-Russian groups rather than by other radical right groups. In 2017, both types of radical right FB groups mentioned Lithuania in the content of their posts the most frequently if compared to the previous years. Lithuania appears 5 302 times in pro-Russian groups and 857 times in radical right groups. Fig. 4 The dynamics of the word Russia in the posts of the radical right groups on Facebook. The word Russia in the topics of pro-Russian groups was most frequently used in 2014 and 2015. This data Fig. 3 The dynamics of the word Lithuania in the posts of the radical right correlate with Russia’s policy and international political groups on Facebook. crises of 2013 and 2015 – after Russian military intervention to Ukraine, various sanctions were imposed on Russia by As was previously mentioned, the increased instances the United States, the European Union (EU) and other of mentioning Lithuania were the most prominent in pro- countries as well as international organizations. In 2015, Russian groups. NATO Strategic Communication Centre of Russia intervened to Syrian civil war (30 September 2015 – Excellence claims that in the period ranging from 1 February 2016) and this event correlates with the dynamics November 2017 to 31 January 2018, the proportion of bot of the topics on Russia in pro-Russian FB groups Russian. activity in Twitter was relatively high, with 62 per cent of all The members of other radical right groups show no attention tweets mentioning NATO and Lithuania [31]. In other radical to this international crisis, the Russian topic in their FB posts right FB groups, Lithuania is mentioned less often as opposed is irrelevant. to the pro-Russian groups. The data in the NATO report correlate with the experimental results. The Russian hybrid TABLE IV. troll or bot activity campaign has reached the users of social The dynamics of the word Russia in the posts of radical right networks in Lithuania, and the experiment shows that this groups on Facebook campaign is still being successfully implemented. According Radica Year to NATO Hybrid trolls (as we have labelled hired, pro- l right 201 201 201 201 201 2015 201 201 Russian trolls), communicate a particular ideology and, most groups 0 1 2 3 4 6 7 Pro- 103 importantly, operate under the direction and orders of a Russian 0 0 0 0 768 2 659 405 particular state or state institution. In the context of the Other 0 3 6 18 57 16 31 47 Ukraine crisis, the aim of hybrid trolls has been to promote Total 104 the Kremlin’s interests and portray Russia as a positive force 0 3 6 18 825 8 690 452 against the ‘rotten West’ and the US hegemony[28]. Russia-related topics seem to be more important to pro- Creating ‘noise’ or ‘informational fog’ around a topic is Russian groups than other radical right FB groups (Fig. 4). a strategy used to distract attention from more strategically The word analysis of the FB groups’ posts that were split to important events. An important example of this has been the 77 case of the downing of Malaysian air flight MH17. Russian In order to compare the content of the posts in radical media channels and social media distributed a large volume right FB groups, a neutral keyword sky was chosen. The of messages offering numerous explanations for why the assessment of dynamics show that the word sky did not plane crashed. Another bot campaign was launched to distract appear in the content produced by the radical right FB groups the public by offering an alternative explanation of the murder of the Russian politician Boris Nemtsov, claiming TABLE VI. that he was killed by jealous Ukrainians. Such ‘news’ were The dynamics of the word sky in the posts of radical right groups published just a few hours after the attack [1]. The experiment on Facebook shows that the word Russia in the pro-Russian groups became Radica Year l right 201 201 201 201 201 201 201 201 more actively used during the turmoil caused by Russia’s groups 0 1 2 3 4 5 6 7 policy. This could have affected the results of the trending Pro- topics in order to make ‘noise’ or ‘informational fog’ around Russian 0 0 0 0 1 5 4 13 Other 0 0 0 0 0 0 0 1 any given topic. Total 0 0 0 0 1 5 4 14 The themes related to land are more relevant to the This indicates that the content generated by the members of both groups. The word count estimations show members of the radical right groups is similar to the political that from 2010 to 2017, the word land appeared 110 times in background. As Veronika Solovian, the administrator of the pro-Russian groups and 141 times in other radical groups popular Finnish-Russian website russia.fi, admits, the trolls “table 5”. The assessment of the thematic dynamics of the are commenting on political topics. They are able to attract groups indicate that in 2014, the word land was more popular other participants into arguments, and other users do not in the posts of other radical right groups than in what was necessarily immediately identify them as trolls [29]. The posted by pro-Russian users (Fig. 5). experiment reveals that political topics are indeed relevant for TABLE V. radical right-wing political groups on Facebook. The largest part of the generated political content could be generated by The dynamics of the word land in the posts of radical right groups trolls or bots. Therefore, social media provides fertile ground on Facebook Radica Year for the dissemination of propaganda and disinformation. The l right 201 201 201 201 201 201 201 201 latter indicates that social media can be an effective tool to groups 0 1 2 3 4 5 6 7 manipulate people’s mind and influence their decisions. Ease Pro- of Use Russian 0 0 0 0 46 15 32 17 Other 0 0 3 41 94 1 1 1 Total 0 0 3 41 140 16 33 18 V. CONCLUSIONS AND FUTURE WORK The word correlates with the Lithuanian land-related Facebook developer acc with API requests and R tools political crisis related to the restrictions imposed on set (Lubridate, Tidytext, ggplot2) can help to analyze radical foreigners who want to purchase land for agricultural right FB groups establishment and themes dynamics. For purposes in Lithuania. The referendum by the Republic of social and political scientists, the most important result is that Lithuania held on 2014 July was related to the in Lithuania radical right groups on Facebook posts together abovementioned restrictions. Prior to the referendum, there with nationalism, strong nation and xenophobic ideology also were many protests and a rally against land purchase appears topics related to the support for the Russian policy restrictions. These events also ignited debates in the virtual and former communist ideology. The analysis reveals that space and affected the topics that were generated in the some radical right groups in addition to the nationalistic radical right FB groups. ideology manifest pro-Russian and pro-socialism ideology. Radical right groups on Facebook started to appear in 2010, but the year 2014 was important for the radical right FB groups as new ideology-following radical right groups appeared and was rapidly growing each year. Experiment data correlates with the awakening of Russian propaganda on social media. The topic word Lithuania is relevant for all the radical right groups. This word in the posts of FB groups appears more than 22 300 times throughout the whole period of groups’ activity. The increased instances of mentioning Lithuania were the most prominent in pro-Russian groups. Russia-related topics seem to be more important to pro- Russian groups than other radical right FB groups and land- related topics is more important to other radical right groups. These topics activity correlates with national or international political crisis: the land-related topics activity reaches its maximum before referendum related to the restrictions imposed on foreigners who want to purchase land for agricultural purposes in Lithuania, the word Russia in the topics of pro-Russian groups was most frequently used in Fig. 5 The dynamics of the word land in the posts of the radical right groups on Facebook 2014 and 2015 while after Russian military intervention to 78 Ukraine, various sanctions were imposed on Russia by [23] J. Silge, D. R.-T. J. of O. S. Software, and undefined 2016, “tidytext: Text mining and analysis using tidy data principles in the United States, the European Union (EU) and other r,” theoj.org. countries as well as international organizations. The [24] H. Wickham, “Simple, Consistent Wrappers for Common String assessment of dynamics show that the word sky did not Operations [R package stringr version 1.4.0].” appear in the content produced by the radical right FB groups. [25] V. Sazonov, H. Molder, and Muur Kristiina, “COMBINED ANALYSIS PREPARED BY THE NATO STRATEGIC This indicates that the content generated by the members of COMMUNICATIONS CENTRE OF EXCELLENCE RUSSIAN the radical right groups is similar to the political background. INFORMATION CAMPAIGN AGAINST THE UKRAINIAN Future plans are to make different kind of radical right STATE AND DEFENCE FORCES,” 2015. groups generated content most frequency words estimations [26] A. Lawrence, “Social Network Analysis Reveals Full Scale of Kremlin’s Twitter Bot Campaign · Global Voices,” 2015. and analyze it dynamics. Future work is also to analyze [Online]. Available: dynamics of FB groups incoming information and the posting https://globalvoices.org/2015/04/02/analyzing-kremlin-twitter- dynamics of most active groups’ members. bots/. [Accessed: 03-Mar-2019]. [27] Svetoka Sanda, “Social Media as a Tool of Hybrid Warfare | VI. REFERENCES StratCom,” 2016. [Online]. Available: https://www.stratcomcoe.org/social-media-tool-hybrid-warfare. [1] B. Podobnik, M. Jusup, D. Kovac, and H. E. Stanley, “Predicting [Accessed: 03-Mar-2019]. the Rise of EU Right-Wing Populism in Response to Unbalanced [28] NATO, “Internet Trolling as a hybrid warfare tool: the case of Immigration,” Complexity, vol. 2017, pp. 1–12, Aug. 2017. Latvia | StratCom,” 2016. [Online]. Available: [2] J. Bartlett, “From hope to hate: how the early internet fed the far https://www.stratcomcoe.org/internet-trolling-hybrid-warfare- right | World news | The Guardian,” 2017. [Online]. Available: tool-case-latvia-0. [Accessed: 03-Mar-2019]. https://www.theguardian.com/world/2017/aug/31/far-right-alt- [29] J. Aro and Y. Kioski, “This is How Pro-Russia Trolls Manipulate right-white-supremacists-rise-online. [Accessed: 03-Mar-2019]. Finns Online – Check the List of Forums Favored by [3] D. R. Forsyth and D. R. Forsyth, Group dynamics, 2nd ed. Propagandists,” 2015. [Online]. Available: Pacific Grove Calif.: Brooks/Cole Pub. Co, 1990. https://www.stopfake.org/en/this-is-how-pro-russia-trolls- [4] NATO strategic communications centre of excellence, manipulate-finns-online-check-the-list-of-forums-favored-by- “Robotrolling 2018/1 | StratCom,” 2018. [Online]. Available: propagandists/. [Accessed: 03-Mar-2019]. https://www.stratcomcoe.org/robotrolling-20181. [Accessed: 03- [30] Wozniak, M., Polap, D., Borowik, G., & Napoli, C. (2015, July). Mar-2019]. A first attempt to cloud-based user verification in distributed [5] “Top 20 Facebook Statistics - Updated March 2019.” [Online]. system. In 2015 Asia-Pacific Conference on Computer Aided Available: https://zephoria.com/top-15-valuable-facebook- System Engineering (pp. 226-231). IEEE. statistics/. [Accessed: 03-Mar-2019]. [6] J. Weng, Y. Yao, E. Leonardi, and F. Lee, “Event Detection in Twitter,” 2011. [7] M. J. Paul and M. Dredze, “You Are What You Tweet: Analyzing Twitter for Public Health.” [8] A. Java, X. Song, T. Finin, and B. Tseng, “Why We Twitter: Understanding Microblogging Usage and Communities,” 2007. [9] A. Ortigosa, J. M. Martín, and R. M. Carro, “Sentiment analysis in Facebook and its application to e-learning,” Comput. Human Behav., 2014. [10] J. Park, W. L.-P. of the fifth international A. conference, and undefined 2011, “Revolution 2.0 in Tunisia and Egypt: Reactions and sentiments in the online world,” icwsm.org. [11] K. Alaimo, “How the Facebook Arabic Page ‘We Are All Khaled Said’ Helped Promote the Egyptian Revolution,” Soc. Media + Soc., vol. 1, no. 2, p. 205630511560485, Sep. 2015. [12] Matt Hicks, “(18) Facebook Tips: What’s the Difference between a Facebook Page and Group? | Facebook,” 2010. [Online]. Available: https://www.facebook.com/notes/facebook/facebook- tips-whats-the-difference-between-a-facebook-page-and- group/324706977130/. [Accessed: 03-Mar-2019]. [13] “Graph API.” [Online]. Available: https://developers.facebook.com/docs/graph-api. [Accessed: 03- Mar-2019]. [14] K. Arzheimer, “Electoral Sociology: Who Votes for the Extreme Right and why-and when?” [15] C. Mudde, The ideology of the extreme right. 2018. [16] J. B.-L. and L. Computing and undefined 1992, “Not unles you ask nicely: The interpretative nexus between analysis and information,” academic.oup.com. [17] D. H.-L. and L. Computing and U. 2004, “Delta prime?,” academic.oup.com. [18] M. Eder, J. R.-L. and L. Computing, and undefined 2012, “Do birds of a feather really flock together, or how to choose training samples for authorship attribution,” academic.oup.com. [19] R. Perera, … S. A.-2010-M. 2010, and undefined 2010, “Twitter analytics: Architecture, tools and analysis,” ieeexplore.ieee.org. [20] Q. Zhao, Y. Tian, Q. He, N. Oliver, … R. J.-P. of the 19th, and undefined 2010, “Communication motifs: a tool to characterize social communications,” dl.acm.org. [21] G. Grolemund, H. W.-J. of S. Software, and undefined 2011, “Dates and times made easy with lubridate,” academia.edu. [22] L. Wilkinson, “ggplot2: Elegant Graphics for Data Analysis by WICKHAM, H.,” Biometrics, vol. 67, no. 2, pp. 678–679, Jun. 2011. 79