How Quantitative Methods Can Inform Qualitative Analysis of Digital Media Discourse Dang-Anh, Mark; Rüdiger, Jan Oliver

www.ssoar.info

From Frequency to Sequence: How Quantitative Methods Can Inform Qualitative Analysis of Digital Media Discourse Dang-Anh, Mark; Rüdiger, Jan Oliver

Veröffentlichungsversion / Published Version Zeitschriftenartikel / journal article

Empfohlene Zitierung / Suggested Citation: Dang-Anh, M., & Rüdiger, J. O. (2015). From Frequency to Sequence: How Quantitative Methods Can Inform Qualitative Analysis of Digital Media Discourse. 10plus1 : Living Linguistics, 1, 57-73. https://nbn-resolving.org/ urn:nbn:de:0168-ssoar-52776-8

Nutzungsbedingungen: Terms of use: Dieser Text wird unter einer Basic Digital Peer Publishing-Lizenz This document is made available under a Basic Digital Peer zur Verfügung gestellt. Nähere Auskünfte zu den DiPP-Lizenzen Publishing Licence. For more Information see: finden Sie hier: http://www.dipp.nrw.de/lizenzen/dppl/service/dppl/ http://www.dipp.nrw.de/lizenzen/dppl/service/dppl/ 57

From Frequency 1. Introduction 2014). As such, they don’t consider the poly- to Sequence: phonic occurrences of language in social me-

How Quantitative n todays’ digitally mediatized world, mi- dia platforms such as Facebook, Twitter, cro-blogging is a common practice in po- Instagram, and so forth. Methods Can Inform This paper aims at bridging the deline- I litical communication (Thimm et al. Qualitative Analysis of 2014). Twitter, as the most popular mi- ated gap between linguistic and medial foci Digital Media Discourse croblogging platform in the western hemi- with an emphasis on the methodological and 1 methodical question of how corpus linguis- (Journal Article + Infographic) sphere , is a medium that is used widely by political actors, be it on an institutional or tics can inform qualitative inquiries in media Mark Dang-Anh | individual level. The course and discourse of linguistics. In other words: Media linguistics Jan Oliver Rüdiger can incorporate quantitative corpus linguis- street protests, especially, is constituted by This paper aims at showing how quanti- microblog communication (Gerbaudo 2012). tics and qualitative hermeneutic analysis tative corpus linguistic analysis can in- Although there are numeral studies on the with respect to the mediality of language form qualitative analysis of digital media role of digital media in street protests, be it use, as every processed occurrence of lan- discourse with respect to the mediality for example in the context of the Occupy guage has its own mediality, i.e. features of of language in use. Using the example of media including the opportunities and con- protest discourse in Twitter, in the field protests (Penney & Dadas 2014), the Tahrir of anti -Islamic ‘Pegida’ demonstrations, a Square protests (Wilson & Dunn 2011; straints they impose on communicative prac- three-step method of collecting, reduc- Tufekci & Wilson 2012), or anti-fascist pro- tices (and vice versa). The case of Twitter ing and interpreting salient data is protests (Dang-Anh & Eble 2013; Neumayer & usage in street protest thus serves as an ex- posed. Each step is aligned with opera- Valtysson 2013), few take a perspective on emplification of how quantitative and quali- tive medial features of the microblog: 2 hashtags, retweets and @-interactions. the linguistic construction of meaning for tative methods might be sensibly combined.

The exemplary analysis reveals the im- and within these streets protests. Further- Given the fact that large numbers of portance of discussions of attendance more, linguistic explorations dealing with people contribute to digital media discourses numbers in protest discourse and the language and protest (cf. Martín Rojo 2014a) on political events, such as protests, and thus asymmetry between administrative (i.e. create large numbers of texts, research the police) and non-administrative dis- focus on on-street/ square semiotics, e.g. on course agents. Furthermore, it exempli- banners (Martín Rojo 2014b), and offline methods must be adapted to these phenom- fies how frequency analysis and se- discourse of street assemblies (Steinberg ena for the purpose of linguistic analysis. quence analysis can be combined for research in media linguistics. 1 Twitter is banned in China. The most popular micro- 2 For illustration, this article is complemented with an This contribution also comprises an blogging platform in China is Sina Weibo. infographic which is available at 10plus1journal.com. infographic which can be retrieved from www.10plus1journal.com. 1 0 p l u s 1 10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics L I v I n g L I n g u I s t I c s

Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 58

However, the sheer vastness of communica- next section introduces Twitter, the opera- the advantage of the quantity of ‘evidencing’ tive occurrences of language in digital media tivity of hashtagging, retweeting and @- data: “by using computer software, analysts does not prevent linguistic analysis from the mentioning and their alignment with the can deal with much larger quantities of data, hermeneutic (re-)construction of meaning proposed three-step method of analysis. and so put forward more convincing evi- through in-depth analysis. Thus, it is our goal Following this, the method is exemplified dence in support of their claims.” (2014: 81) to show how corpus linguistics as a quantita- using the case of ‘Pegida’ protests and a cor- Both views emphasize the utility of corpus tive means can heuristically inform qualita- responding Twitter corpus. We conclude linguistic methods for the analysis of large tive analysis of digital media discourse. To be both, the methodological reflection and the data sets. However, data and analyses based more precise, we will show how the analysis exemplary analysis, in the last section. on data, be it from small or large corpora, of frequencies structures and filters large must always be interpreted and contextual- datasets in order to practically perform qual- 2. Corpus Linguistics as Heuristics for ized in order to conduct an adequate recon- itative analysis of communicative sequences. Qualitative Analysis structive analysis of the construction of By identifying salient utterances and speak- meaning within social contexts. From such a ers through quantitative measures the perspective, evidentiality4 or cogency is not an When it comes to researching large data three-step method of data collection, reduc- inherent feature of a datum but evidencing is sets, it is increasingly regarded as sensible tion, and interpretation points to a commu- a reflexive scientific process that data ana- and fruitful to extend qualitative methods by nicative sequence, the meaning of which is lysts negotiate intersubjectively. As a conse- quantitative means (cf. O'Halloran 2010; qualitatively analysed with respect to its quence, quantitative and qualitative meth- O'Keeffe 2012; Bubenhofer 2013; Baker & context. However, our approach is not strict- ods have to be balanced precisely. While Levon 2015). McEnery & Hardie, thus taking ly linear but circular, as we reassess the qualitative methods still pave the royal road a stance on corpus linguistics as a supportive quantitative analytical steps by corpus- to reconstructive analysis of language-in- extension of qualitative research methods: hermeneutic reflections.3 use, quantitative methods might support “any field that is based, primarily or in part, The following section outlines the rela- analyses heuristically. By accessing large on the study of text can benefit from corpus tion between quantitative and qualitative data sets heuristically with the use of corpus methods in any research context where the methods with special regard to the notion of linguistics, subtle communicative patterns body of text that is of interest expands be- salience in conjunction with those of fre- might emerge, intuitively expected commu- yond the point where hand-and-eye meth- quency and ascriptions of relevance. The ods of analysis can fully encompass its con-

tents.” (2012: 231) In their work on written 4 Evidentiality is not used here in its specific linguistic 3 This is exemplified in the first step of the exemplary discourse, Cameron & Panović emphasize terminological sense that points to grammatical fea- analysis (cf. p.64). tures (cf. Chafe & Nichols 1986).

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 59 nicative patterns might be confirmed and visible acts of political communication that 2) “a politically relevant topic (under the thus the foci of qualitative analyses can be characterize political systems or coin politi- conditions of a democratic system, this directed to distinct pieces of data.5 cal discourse over a trans-situational time predominatly means a controversial span. However, Klein acknowledges that topic)” 3. Salience and Frequency salient sentences necessarily belong to a 3) “a special situation that is evoked by collective actual knowledge and possibly public attention, aggravation, or arous- We use salience as a key concept to identify anchor within the collective memory of polit- ing a latent controversy”. distinct communicative sequences from ically and historically interested parts of so- large data sets. Klein (2014) conceptualizes a ciety (Klein 2014: 122). From Klein’s (2014: As the first two – rather vaguely attributed – trans-situational understanding of “salient 123) point of view, three features are re- features of considerability and political rele- sentences” by which he refers to epoch- quired for sentences to become salient: vance indicate, Klein clearly has in mind defining utterances6 such as the first sen- larger political issues with their own historic- tence of the Basic Law for the Federal Re- 1) “a considerable speaker (person, group- ity. Contrary to such discourses mainly driv- public of Germany (“Human dignity shall be ings)” en by mass media, in digital media discourse, inviolable.”; Basic Law for the Federal Re- the first two requirements are not “indispen- public of Germany 2012: 15) or sentences sable” (Klein 2014: 123), as speakers and that provoke political discussion, such as schichte. Aber der Islam gehört inzwischen auch zu topics might emerge as well, i.e. postings Germany’s former president Wulff’s utter- Deutschland.“ In English (our translation): “Christian- from non-prominent speakers about non- ity belongs without any doubt to Germany. Judaism current topics might trigger public debates. ance in a speech: “Islam belongs to Germa- belongs without any doubt to Germany. But Islam ny.”7 Salience, in this sense, points to highly meanwhile also belongs to Germany.” The translated However, even for singular events – such as manuscript from his speech on Reunification Day in those discussed here that occurred in a se- Bremen, 3rd October 2010 is available on the Ger- 5 However, this merely describes one way amongst man parliament’s website (cf. Wulff 2010). Interest- ries of iterative protest phenomena under others to cherry-pick the focus of qualitative text ingly, the official translation adds the notion of ‘iden- the label of PEGIDA – the notion of salience analyses (cf. Baker & Levon 2015). tity’ to Wulff’s utterance: “Christianity is without a 6 is helpful for our approach to identify rele- As sentence is a categorization mainly for written doubt part of German identity. Judaism is without a language, we prefer the term utterance from a prag- doubt part of German identity. Such is our Judaeo- vant speakers and utterances within a cor- matical and media linguistic point of view in order to Christian heritage. But Islam has now also become pus of thousands of tokens. In a broader overcome the spoken-written-dichotomy and to part of German identity.” These nuanced transla- prevent from a “written language bias” (Linell 1982). tional differences create differences in meaning that sense that emphasizes the aspect of percep- 7 This reproduction is slightly shortened. Wulff origi- cannot be further discussed here. However, the de- tivity from the perspective of mass commu- nally uttered: “Das Christentum gehört zweifelsfrei liberate defusing of the original utterance by official nication research, salience is understood as zu Deutschland. Das Judentum gehört zweifelsfrei zu authorities underscores the salience of Wulff’s ut- Deutschland. Das ist unsere christlich-jüdische Ge- terance in Klein’s sense. the act of “making a piece of information

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 60 more noticeable, meaningful, or memorable The understanding of utterances that are ascribe relevance to postings by singular to audiences. An increase in salience en- frequently made relevant (and thus become acts of retweeting, the accumulation of re- hances the probability that receivers will salient) in the course of their interpretative tweets makes this relevance more perceiva- perceive the information […]” (Entman reception yields methodological implications ble – as indicated by the retweet count un- 1993: 53). Despite arguing with the most for the field of salient sentences: “For its der each posting – and as such salient. Sali- problematic concepts from (mass) communi- systematic processing the conjunction of ence is thus established by reflexive “quanti- cation studies like audience and receiver, one corpus linguistics and linguistic hermeneu- fiable valuation practices” (Paßmann 2015: can transfer the concept of salience to Twit- tics might be the appropriate methodical 141) that social media platforms make poster, where the relation between the original approach.” (Klein 2014: 125).9 Consequent- sible through their medial features. As a con- poster, the retweeter and the recipient (of ly, computational procedures of corpus lin- sequence, the visibility of practices in social the retweet) is cascading as the latter can guistics and intersubjectively scrutinized media platforms makes them accessible for herself become a retweeter (and so forth). qualitative-hermeneutical methods should research: “Platform activities [such as re- As such, making a posting relevant by re- not be viewed as contradictory but as com- tweets] directly connect practices with data tweeting it is a productive practice whereas plementary (Felder 2012: 125). that are generated in the process of retweet- perceiving relevance and thus the feature of In social media platforms, contrary to ing. Thus, user activities become summable.” a posting as salient is a merely receptive one. the production of resonance by gatekeepers (Paßmann & Gerlitz 2014: 2)11 Taking into Consequentially, Klein identifies reso- in mass media discourse, it is the recipients account the reflexivity of social media data nance as the key factor of responsive ascrip- who make other agent’s postings more rele- (Paßmann 2014) and the agency of social tions of salience to utterances: vant by retweeting them.10 Whereas agents media users that finds expression in the communicative practices of social media Whether a sentence is apprehended, spread hängt aber von der Resonanz bei den Rezipienten ab, users, the categories of considerable speakers and finally becomes commonly known depends insbesondere bei den politischen Leitmedien (die großen TV-Sender, überregionale Zeitungen und po- and politically relevant topics in social media on the resonance amongst the recipients, espe- litische Magazine in Print- und Online-Format). Da- discourse are not historical in Klein’s sense cially leading political media […]. Here, quantity bei spielen Quantität (mediale Distribution) und In- (medial distribution) and intensity (degree of tensität (Grad der Hervorhebung, Zitieren oder Re- ferieren) ein zentrale Rolle. (Klein 2014: 123) Anh et al. 2013b) and thus platforms like Twitter Inc. accentuation, citing or referring) play a central 9 Original: “Für seine systematische Bearbeitung dürf- (Halavais 2014) have their own agenda, doing plat- 8 role (Klein 2014: 123). te die Verknüpfung von Korpuslinguistik und linguis- form politics (Gillespie 2010). tischer Hermeneutik der angemessene methodische 11 Original: “Indem Plattformaktivitäten eine direkte Ansatz sein.“ (Klein 2014: 125) Verbindung zwischen Praktiken und den dabei er- 8 In this paper, all German citations are translated into 10 It is important to note that recent studies in media zeugten Daten herstellen, werden Daten von Nut- English. Original: „Ob ein Satz aufgegriffen, weiter- research point to the fact that the distribution of zeraktivitäten aggregierbar.“ (Paßmann and Gerlitz getragen und schließlich allgemein bekannt wird, postings in social media platforms is selective (Dang- 2014: 2).

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 61 but are ascribed perceivable relevance and, Assuming that “the most retweeted ac- Twitter is especially useful for time-sensitive as such, salience is established through user counts represent key agents of the protest digital communication during dynamic practices. discourse” (Dang-Anh & Eble 2013: 2; cf. events such as street protests. At the time of In Twitter, salience emerges from the Wilson & Dunn 2011) one should identify writing, postings can have four distinct oper- frequent and perceivable communicative these key agents as well as their postings. ative in-text features: hashtags, @-mentions, practice of ascribing relevance by retweet- retweets and hyperlinks. ing. From such a user-centered perspective, 4. Twitter The role Twitter plays in street protests the formerly blurred genesis of relevance ranges from supportive to constitutive. With becomes clearly identifiable, and even dis- Twitter is, by far, the most popular mi- the help of Twitter, protestors and observers tinctively traceable, through the analysis of croblogging platform. The central linguistic are able to organize and coordinate protest, retweet frequencies. Consequentially, “the unit of analysis is the posting, called tweet in inform themselves about what is going on in corpus processes drive the analysis, and lin- everyday language. However, the term tweet the streets, mobilize and navigate to rele- guistic patterns based around what emerges was coined by Twitter Inc. for marketing vant places (Dang-Anh & Eble 2013), evalu- as frequent or salient need to be accounted purposes. Thus, we use the term posting as a ate situations and political actions, support for.” (Baker & Levon 2015: 222) Salience synonym due to its unrelatedness to specific each other by expressing solidarity, insult or then is not only perceivable on the front-end social media platforms: monitor political enemies, build interperson- – which is the case for salient retweets but al relationships and strengthen ties amongst not for the salience of frequently uttered We introduced the category posting as a themselves, inform themselves about the words or phrases – but also becomes detect- basic element to capture CMC micro- and course of the protests, celebrate or regret able by frequency analysis. In Twitter, two macrostructures. A posting is defined as a the outcomes or course of a protest event, objects of analysis might be distinguished content unit that is being sent to the server comment on the protest and so forth (cf. when detecting retweet frequencies: ‘en bloc’. Postings can usually be recognized Penney & Dadas 2014). These and many by their formal structure, even if they have more communicative practices are predomi- different forms and structures across CMC 1. the most retweeted tweet and nantly performed in Twitter by using four genres. This facilitates the automatic seg- 12 2. the most retweeted account. mentation and annotation of CMC micro- and operators: hashtags, retweets, @-mentions and hyperlinks. macrostructures. (Beißwenger et al. 2012: 5) 12 In their study of hashtags in political communication, Barash and Kelly correspondingly focus on With its restriction to 140 characters per “Peakedness, which measures the broad appeal and salience of a contagious phenomenon”(Barash & posting and its specific distribution features, Kelly 2012: 5).

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 62

4.1 The Operativity of Signs in Digital include: hyperlinking, including the change retweet button, ‘#’, and ‘@’ and give a brief Communication of the font color and underlining, sending description of their operativity, as these op- user notifications, republishing postings and erators “provide important parameters for The notion of operativity refers to character- many more. quantitative evaluation and qualitative in- istics of digital communication platforms and In Twitter, the executable operations terpretation” (Klemm & Michel 2014: 95).15 denotes non-human operations that are in- are processed by the practices of hash- structed by human beings through the use of tagging, retweeting, @-mentioning and hy- 4.2 Hashtags distinct operative characters or buttons. It perlinking whereas the computational oper- stems from the concept of the (auto-) opera- ation coincides with the corresponding Hashtags are used to structure discourses tivity of script. Grube (2005) distinguishes communicative acts, e.g. citing somebody by and discourse fragments in Twitter and thus between referential script that is assigned to retweeting her posting (cf. Thimm et al. enhance the visibility of tweets (Page 2012). the “writing systems of our narrative and 2011). As for our purpose, we will concen- People contribute to a specific topic by using reasoning textual forms” (Grube 2005: 81), trate on the first three functionalities initiat- a specific hashtag as “’inline’ metadata” that operative script that “underlies the cultural ed by the usage of the operators ‘RT’14 or the makes topically related tweets searchable technique of written calculation” (Grube and findable (Zappavigna 2011: 791). Fur- 2005: 81) and auto-operative script, “in processes that lie below the perceptional surface of thermore, “Twitter users frequently create communication media into consideration. Computer- which single signs of a [notational] system mediated communication is an everyday business of idiosyncratic hashtags to add a layer of aren’t manipulated [by manual operations delegating complex computational processes by sim- meaning to a word or phrase” (Dayter 2015: ple instructive practices, such as pressing buttons or like written calculation] but automatically using operators, without detailed knowledge of 6). In both respects, hashtags contextualise processed” (Grube 2005: 82). Operativity in these computational processes. An analogue logic is digital communication means that, through stated by Bishop Berkeley in 1707 for mathematic operations: “The rules may be practiced by men who the use of operative signs – operators –, au- neither attend to, nor perhaps know the principles … or commenting on an initial posting differs from re- thors instruct the machine to execute opera- and as any ordinary man may solve divers numerical tweeting by using the retweet button (Paßmann & 13 questions by the vulgar operations of arithmetic, Gerlitz 2014). In the meantime, Twitter has intro- tions (Grube 2005: 82). These operations which he performs and applies without knowing the duced a new function that allows users to add anoth- reason of them: Even so… you may operate, compute er 140 characters to a button-initiated retweet (cf. and solve problems thereby, not only without an ac- Parkinson 2015). 13 The reason for the success of operative digital com- tual attention to, or an actual knowledge of the 15 Original: “Möglich machen dies neuartige Vernet- munication media is its instructional relation be- grounds of method.” (qtd. in Krämer 1991: 123-124). zungsstrukturen über technische Operatoren wie tween humans and machines: by the rather simple 14 The practice of manually adding ‘RT’ to tweets is not RT, # oder @ – die zugleich für die Forschung wichti- use of operators, agents delegate complex tasks to an automated operation but using the retweet but- ge Parameter für die quantitative Auswertung wie the machine and its (black-boxed) algorithmic code. ton is. Paßmann & Gerlitz point out that the practice qualitative Interpretation bereitstellen.” (Klemm & Media linguistics should take such infrastructural of retweeting by typing “RT” and potentially editing Michel 2014: 95)

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 63 utterances and thus function as contextuali- relevance of accounts is the number of fol- 5. Three-Step Method sation cues (Dang-Anh et al. 2013a). lowers (Paßmann 2014), another is the The former specification and thus the selec- amount of retweets of an account’s postings. For the sensible combination of quantitative tion of messages can be further differentiat- This widens the range of a particular posting, and qualitative analyses, we propose a ed by the use of subordinating hashtags. The not only at one point in time but, additional- three-step process of collecting, reducing Twitter discourse of #PEGIDA, for example, ly, each and every time a posting is redistrib- and interpreting16 data. In our case, we com- does not only relate to the anti-Islamic pro- uted by retweeting. Thus, the amount of re- bine corpus linguistic frequency analysis tests in the city of Dresden, where the pro- tweets of a protest-related message with sequence analysis that has its origins in tests emerged, but also to those that hap- throughout the course of a protest event is ethnomethodological conversation analysis pened in other cities throughout Germany. an important measure of relevance whereas (cf. Bergmann 1981). Further differentiation of discourse was thus the salience of postings and accounts is, as done by using other place-related hashtags stated above, determined by frequency 1. Data collection. It is crucial to identify like #MAGIDA for anti-Islamic protest in analysis. significant terms and entities as well as Magdeburg or #FRAGIDA for those in Frank- an appropriate time span for the seg- furt am Main, just to name two of the widely 4.4 @-mentions ment of discourse that is to be studied. dispersed wave of anti-Islamic protests in On this basis, the corpus can be com- Germany in the winter of 2014/15. Counter- The @-operator is used to directly address piled accounting for the discourse struc- protest was often tagged with the prefix “no- (@-adressing) accounts or to mention (@- turing practices of the agents them- “, e.g. #NOPEGIDA, #NOMAGIDA or mentioning) them within the text of a post- selves.17 #NOFRAGIDA. ing. Depending on the clients’ preferences, the addressee might receive a message

4.3 Retweets about his being addressed and thus takes 16 It is vital to note though, that interpretation comes note of it. This may lead to a multiple-turn into play at every stage of the research process. 17 However, there are limitations on collecting data Retweets are used to redistribute and dis- interaction, though, interactions on Twitter from social media platforms. It is important to note seminate others’ postings, to comment on seldom consist of more than an initial tweet- that any selection of identifying linguistic and tem- them, to publicly agree with them, to cite response-structure (cf. Honeycutt & Herring poral criteria for data collection involves omitting other data. In social media platforms, access to data people’s utterances, thus making certain 2009). In protests, @-addressing is often is restricted in many ways (cf. boyd & Crawford tweets and Twitter accounts relevant (boyd used towards relevant accounts, measured 2012, Puschmann & Burgess 2014). Therefore, it often is very hard, if not impossible, to determine a et al. 2010). Whereas one measure of the by the quantitative means mentioned above. basic population for Twitter or Facebook. Additional- ly, users of social media, and this is especially true for

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 64

2. Data reduction. Distinct salient agents tional and transsituational contexts and (Köln), BRAGIDA in Braunschweig, MAGIDA and utterances can be detected by fre- background knowledge of the research- in Magdeburg or BAGIDA in Bavaria, just to quency analysis. Any corpus linguistic ers that is methodically substantiated name a few. Soon after the first assemblies method ascertaining frequencies of based on the intersubjective insights of PEGIDA in Dresden, counter-protests words, phrases, n-grams and significant from hermeneutic-interpretative nego- were established under the label of co-occurrences etc. might be suitable, tiations.18 ‘NOPEGIDA.19 depending on your research interest. In social media, textual data is linked with 6. An Exemplary Analysis of a Twitter 6.1 Step 1: Hashtags – Collecting metadata that allows for nexuses of lin- Dataset on #Fragida Data Based on Discourse- guistic phenomena such as time, au- Relevant Hashtags thors, connections between authors, What is #Pegida and #Fragida? connections between postings, profile The corpus20 data was collected from the pictures, trans- and intermedial links Germany faced a wave of right-wing anti- Twitter-Stream-API21 from 30th January and so forth. For our purpose of linguis- Islamic protests under the label of PEGIDA 2015, 0:00:00h to 4th February, 23:59:59h. tic frequency and sequence analysis, the (Patriotic Europeans against the Islamisation of Every posting that contained selected que- nexus of text, author and time is rele- the occident) in winter 2014/2015 (Daphi et ries22 and hashtags identifying the discourse vant. al. 2015). The protests originated in Dresden 3. Data interpretation. Qualitative analysis in October 2014 and had several subsidiar- 19 must be performed with regard to the Whereas NOPEGIDA is a loose label, the two main ies in other German cities. The abbreviation actors organising counter-protests in Dresden were sequentiality and situatedness of the PEGIDA was mostly altered with the city’s or Dresden für alle (Dresden for everybody) and Dresden communicative occurrences. Bringing region’s names, e.g. KÖGIDA in Cologne nazifrei (Dresden free of Nazis). However, for both labels of NOPEGIDA (2015) and NOFRAGIDA interactions into sequence reveals in- (2015) from Frankfurt, there are Facebook groups teractive phenomena with respect to 18 Sequence analysis in digital media, however, is dif- with several thousand followers. 20 their characteristics of being jointly ferent from conversation analysis of talk-in- Unfortunately, Twitter prohibits making the corpus interaction due to its disrupted sequentiality, e.g. in publicly available for further research (cf. Pusch- constructed. As an act of reflexive con- Twitter, turn sequences are not linear and distinctly mann & Burgess 2014). Although this contradicts our textualisation, hermeneutic qualitative distributed from account to account. Beißwenger research attitude, we defer to these restrictions due and Storrer (2008: 301) note: “due to the technical to legal considerations. analyses draw on the cotext, the situa- (not pragmatic) sequencing of participant submis- 21 For more details on collecting data from Twitter cf. sions […], there is a larger margin for interpretation Gaffney & Puschmann (2014). (or speculation) in the modelling of CMC data than in 22 Search terms were: BAGIDA, BOGIDA, BRAGIDA, Twitter, only represent a “very particular sub-set” the modelling of data from (oral/face-to-face) con- DAGIDA, DUEGIDA, DÜGIDA, DUIGIDA, FRAGIDA, (boyd & Crawford 2012: 669). versations.” HOYGIDA, KAGIDA, KOEGIDA, KÖGIDA, LEGIDA,

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 65 on PEGIDA and its subsidiaries was collected. 6.2 Step 2: Retweets – Reducing Data collection and frequency analysis were Data Based on Quantitative performed with the tool CorpusExplorer.23 Frequency Analysis of Retweets Data cleansing included eliminating dupli- cates and spam, and filtering for German The most retweeted tweet was posted by tweets. Further data processing comprised the account @Polizei_Ffm on 2nd February, lemmatising and tokenising. The corpus con- 2015 at 7:55 pm: sists of 50,633 postings sent by a total of 14,950 accounts and containing 1.06 million 2015- Here the official attendance 24 numbers from tonight: #Pegida/ tokens of 55,834 types. 8 02-02 #Fragida: 85 #nofragida: 1200 We detected an unexpectedly high 19:55 #meinfrankfurt #Hauptwache. amount of tweets in the context of the

PEGIDA protests in Vienna. Qualitative re- It is the “official account of the police Frank- 27 viewing of these tweets revealed a high ratio Figure 1: Screenshot of the most retweeted tweet. furt am Main during operations” as stated in of postings from news agencies and media 25 the profile. In May 2015, the account ap- displayed below the text in the screenshot accounts that were not directly related to proved by Twitter had about 20,000 follow- (Figure 1). The posting consists of four lines the street protests. Therefore, we decided to ers. During the data collection period, it was that are separated by line breaks. The first disregard the corresponding Vienna tweets. 26 retweeted 94 times and faved 85 times, as line “Here the official attendance numbers Hence, through this corpus-hermeneutic from tonight:” announces the attendance analytical step, quantitative analysis could nd 24 numbers of the protests that occurred on 2 be reassessed by qualitative means circular- This is one of several tweets analyzed in the sequence analysis. The columns from left to right con- February, 2015 in the city of Frankfurt am ly. tain: the postings’ index, the posting time, the posting text. The whole list of postings is attached in the ap- Main. The second line announces the num- pendix. bers of the attendees of the right-wing anti- 25 Cf. Polizei_Ffm (2015). As Paßmann (2015: 156) Islamic protest. The hashtags #PEGIDA and remarks, the white tick in a blue circle (cf. Figure), which marks an account as a verified account being #FRAGIDA are used to indicate the reference identity-proofed by Twitter, indicates a top-down as- object of 85 protest attendees. Whereas MAGIDA, MUEGIDA, MÜGIDA, NUEGIDA, sessment by the platform corporation Twitter Inc. NÜGIDA, OGIDA, PEGIDA, ROGIDA. whereas assessments like the follower count are bot- 23 CorpusExplorer was programmed by one of the tom-up. 27 This screenshot has been edited: the profile pictures authors, Jan Oliver Rüdiger. It is available as open 26 For the practice of faving cf. Paßmann & Gerlitz of the retweeting account have been eliminated for source software on www.corpusexplorer.de. (2014). anonymization.

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 66

#PEGIDA functions as a collective label for the posting. However, from decontextual- counter-protestors’ assembly.28 It states the right-wing anti-Islamic protests that ized analysis of the single posting, it is hard that more than 300 people had yet attended were going on throughout Germany at the to judge the reason for its ascribed rele- protests at Hauptwache, a square in Frank- time, #FRAGIDA marks the local reference to vance. Thus, it is necessary to contextualize furt where the assemblies took place and the city of Frankfurt. For the opposition, the the posting further by looking at the negotia- that is shown in the photo posted with the counter-protests are labelled as tions that the posting evokes and that it is tweet. #NOFRAGIDA and the Frankfurt police embedded in. counts 1200 attendees in the third line of 2015- More than 300 attendees are at the #Hauptwache already in the posting. The last line contains two 6.3 Step 3: @-mentions – Interpret- 1 02-02 #meinFrankfurt #nofragida hashtags, #meinfrankfurt, under which city- ing Data Based on Sequence 16:50 #fragida [URL photo] related tweets are posted continually and Analysis of @-interactions

#Hauptwache that refers to a building and Whereas in this posting @Polizei_Ffm does square in the city. Sequence analysis in Twitter necessitates not distinguish between the protests of The high amount of retweets indicates putting postings reconstructively and selec- FRAGIDA and the counter-protests that Twitter users ascribe relevance to this tively in a reasonable, coherent order. For NOFRAGIDA, it does so in further postings. In utterance and thus salience that not only this purpose and following the reduction of the course of the protests, @Polizei_Ffm becomes visible by frequency analysis but the second step, our analysis focuses reports on the attendance numbers three also by the retweet index under the text. As more times, including both protests, an administrative authority, the police was in a) on the postings of the most retweeted FRAGIDA and NOFRAGIDA. operation in the scope of the anti-Islamic account (@Polizei_Ffm) that are topical- protests and the counter-protests in Frank- ly related to the most retweeted tweet Intermediate status attendance nd 2015- furt on 2 February, 2015. Not only because (8) and numbers: #Pegida/ #fragida: 50 2 02-02 #nofragida: 620 #meinfrankfurt of this prominent societal role of the account b) on the interactions that are technically 17:37 but also due to the use of the protest-related (3,4,5) or topically (9) linked to the these #Hauptwache [URL photo] hashtags of both political opponents postings (cf. Table 1 in the Appendix). 2015- Intermediate status attendance numbers: #Pegida/ #fragida: 60 FRAGIDA and NOFRAGIDA, the tweet might 7 02-02 #nofragida: 1180 #meinfrankfurt have gained high visibility amongst protes- 18:42 The first posting regarding the attendance #Hauptwache [URL photo] tors and observers. Additionally, it thematis- numbers was posted by @Polizei_Ffm at es the topic of protest attendance numbers, 16:50, 20 minutes after the start time of the which is another reason for the relevance of 28 Cf. NOFRAGIDA (2015).

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 67

Here the official attendance num- ed above: an evidencing process – is doubt- 2015- 2015- The @Polizei_Ffm has counted: bers from tonight: #Pegida/ 8 02-02 9 02-02 85 #Pegida-demonstrants op- ed. The hashtag #DankePolizei (‘thank you, #Fragida: 85 #nofragida: 1200 posed 1200 counter- 19:55 20:14 police’) is conventionally used ironically for #meinfrankfurt #Hauptwache. demonstrators. postings reporting negatively on police, e.g. cases of police violence. As such, it marks the In postings 2 and 7, @Polizei_Ffm chooses an It alleges an evidencing process of counting identical wording of the posting with chang- although @Polizei_Ffm has not stated such a critical stance towards police authorities. Furthermore, the posting offers an alterna- ing attendance numbers. “Intermediate sta- process in posting 8. tus” here marks a snapshot in the course of Looking at the interactions occurring tive second assessment on attendance numbers as an adjustment of the first assessment the protests and presupposes the anticipa- around the theme of attendance numbers, tion of a rise in attendees. In posting 8, the negotiation processes occur: (Pomerantz 1984). Postings 4 and 5 ask iron- ic questions on the whereabouts of FRAGIDA most retweeted tweet, “official attendance attendees. This presupposes an alternate numbers” are reported. In contrast to post- Intermediate status attendance 2015- perception on-site and, as such, the posters’ ings 2 and 7, posting 8 constitutes a final 2 02-02 numbers: #Pegida/ #fragida: 50 #nofragida: 620 #meinfrankfurt statement regarding the attendance num- 17:37 physical presence. In posting 6, @Polizei_Ffm #Hauptwache [URL photo] revises their first estimation of 50 FRAGIDA bers. “Official” here claims to state a reliable 2015- . @Polizei_Ffm The #DankePolizei attendees and downgrades the number to fact and, as such, interpretational sovereign- 3 02-02 cannot count: 15 Fragida on the ty over the situation, based on the institu- 17:40 spot ..... 30. Commenced with an excuse, the police tional character of @Polizei_Ffm as an ad- 2015- concede misjudgment of attendance num- 4 02-02 @Polizei_Ffm Where are those 50 ministrative agent. Such a turn harnesses the hiding? #nofragida bers. “Obviously” here refers to the congru- 17:43 ent commentaries of the respondents on the institutional asymmetry between authorities 2015- .@Polizei_Ffm Are the other 47 in and non-administrative agents for the (re-) 5 02-02 divergent on-site perceptions. Only in this the Katharinenkirche just now? production of an asymmetry of knowledge 17:45 posting, does @Polizei_Ffm refer to the act of counting as an evidencing practice to em- (Rintel et al. 2013). Consequentially, the as- 2015- . @Polizei_Ffm Sorry, the #Pegida 6 02-02 number was too high, obviously. phasize their adjustment of attendance sessment of the protest situation is adopted Currently, the colleagues have 17:53 numbers and finally thank the respondents by news media representatives, such as the counted 30. Thanks for the pointer. for their allusions. sender of the following posting, RTL Hessen:

In postings 3-5, as responses to the initial

posting 2 by @Polizei_Ffm, disagreement is expressed. In 3, the ability to count – as stat-

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 68

6.4 Discussion tions. As such, they assign opposing parties 7. Conclusion to places and separate them. Thus, they have The analysis provided is exemplary and thus privileged physical access to protest sites (cf. In alignment with the operative mediality of abridged. However, it provides initial in- 1 and 2) thus a structural asymmetry of Twitter, we have proposed a simple three- sights on the major importance of attend- knowledge (Günthner & Luckmann 1995) is step method in order to heuristically focus ance numbers estimations and reportings for inherent for political protest. As the interac- from big (amounts of) data to salient practic- the assessment of political protest. First as- tions reveal, physical access and thus the es of communication that can be qualitative- sessments here were stated by police au- ability to visually perceive protest sites is a ly analysed. Firstly, for collecting data, it is thorities and controversially negotiated. precondition for assessments on attendance crucial to identify pertinent terms and enti- Raymond & Heritage (2006) identify “three numbers. However, the respondents (post- ties as well as an appropriate time span for features of assessment sequences” that “are ings 3-5) implicitly claim to somehow have the segment of discourse that is to be stud- especially relevant for such negotiations” visible access. The asymmetry regarding ied. For our exemplary Twitter analysis, we (Raymond & Heritage 2006: 684) and can be access is, for the time being, reversed by the have exploratively identified hashtags and applied to protest communication in micro- police’s comment on having (re-)counted the words that mark the PEGIDA discourse. Sec- blogs: attendance numbers. By posting the “offi- ondly, for reducing data, distinct salient cial” attendance numbers in the most re- 1) “in order to offer assessments of states of agents and utterances can be detected by tweeted posting, however, asymmetry is affairs, and so in order to agree or disa- frequency analysis. We selected retweets restored. gree, parties must have some access to that indicate the relevance-making practices Additionally, asymmetry is also estab- them.” of the involved agents. Finally, for interpret- lished by the mediality of Twitter: “highly 2) “speakers rank their access to whatever is ing data, qualitative analysis must be per- visible users determine what gets amplified being evaluated” formed with regard to the sequentiality and and what does not. Twitter’s reality is one of 3) “offering a first assessment carries an situatedness of the communicative occur- asymmetric visibility.” (Fuchs 2014: 192). implied claim that the speaker has prima- rences. In our case, we have chosen sequen- Thus, social relations established by and ry rights to evaluate the matter assessed” tial analysis as a means for (re-)constructing through medial features within a historical (Raymond & Heritage 2006: 684). the interactional negotiations of meaningful process of social and communicative prac- contributions, which are initiated via @- For the police’s postings, all three aspects tices (of relating to each other, e.g. by follow- mentions, to protest discourse. However, pertain. Holding the monopoly on the legiti- ing someone) predetermine asymmetric re- any qualitative hermeneutic-interpretative mate use of force, the police is responsible lations that themselves reproduce asymme- method might be applied to salient utteranc- for the peaceful execution of demonstra- tries amongst social media users.

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 69 es. Our focus was to describe how corpus practices of making-topics-relevant consti- Basic Law for the Federal Republic of Germany, linguistics can be helpful as a heuristic tool tute countable as well as assignable indica- Deutscher Bundestag Art.1 (2012). Beißwenger, M. et al. (2012). DeRiK: A German Ref- for qualitative analysis of digital media dis- tors of relevance and, as such, salience. erence Corpus of Computer-Mediated Communica- course in media linguistic inquiries. Hence, salient communicative practices in tion. Paper presented at the Digital Humanities The internet renders visible a “norma- digital media are traceable and thus become 2012 Conference. Beißwenger, M. & Storrer, A. (2008). Corpora in tive interactive social order” (Thielmann key data for the analysis of social processes Computer-Mediated Communication. In A. Lü- 2012: 101), i.e. it reproduces an asymmet- such as protest events. However, it is vital to deling & M. Kytö (Eds.), Handbücher zur Sprach- rical relation between administrative and note that these kinds of analyses focus on und Kommunikationswissenschaft = Handbooks of Linguistics and Communication Science: HSK 29.1. non-administrative agents that negotiate communicative practices in the scope of pro- Corpus Linguistics: An International Handbook political issues. Asymmetry, in the analysed test events – nothing more, nothing less. In (pp. 292-309). Berlin, New York: Mouton de case, is explicated by the police displaying order to achieve deeper levels of under- Gruyter. Bergmann, J. (1981). Ethnomethodolgische Konver- the reported numbers of demonstration par- standing and contextualization of communi- sationsanalyse. In P. Schröder & H. Steger (Eds.), ticipants as ‘official’ and is amplified by nu- cative practices in digital media, it might be Dialogforschung. Jahrbuch 1980 des Instituts für merous retweets. Thus, there is both an advisable to complement future research deutsche Sprache (pp. 9-51). Düsseldorf: Pädago- gischer Verlag Schwann. asymmetry of reputation, regarding the fol- designs with additional methods, particularly boyd, d. & Crawford, K. (2012). CRITICAL QUES- lower counts, and a difference between ad- with ethnographic approaches (cf. Androut- TIONS FOR BIG DATA: Provocations for a Cul- ministrative and non-administrative agents sopoulos 2008). tural, Technological, and Scholarly Phenomenon. Information, Communication & Society, 15.5, pp. as well as an asymmetry of knowledge which 662-679. then is negotiated. By assessing and negoti- References boyd, d. et al. (2010). Tweet, Tweet, Retweet: Con- ating the number of protest participants versational Aspects of Retweeting on Twitter. In Proceedings of HICSS-43 (pp. 1‐10). with the use of @-mentions, the interlocu- Androutsopoulos, J. (2008). Potentials and Limita- Bubenhofer, N. (2013). Quantitativ informierte tors not only evaluate political protest tions of Discourse-Centered Online Ethnogra- qualitative Diskursanalyse: Korpuslinguistische phy. Language@Internet, 5. events but make the topic of participants’ Zugänge zu Einzeltexten und Serien. In K. S. Baker, P. & Levon, E. (2015). Picking the Right Roth & C. Spiegel (Eds.), Angewandte Diskurs- numbers relevant for protest discourse. Such Cherries? A Comparison of Corpus-Based and linguistik. Felder, Probleme, Perspektiven (pp. 109- assessments and negotiation processes – in Qualitative Analyses of News Articles about 134). Berlin: Akademie Verlag. Masculinity. Discourse & Communication, 9.2, pp. what Thielmann refers to as accountable Cameron, D. & Panović, I. (2014). Working with Writ- 221-236. ten Discourse. Los Angeles: SAGE. social media, i.e. attributable, quantifiable, Barash, V. & Kelly, J. (2012). Salience vs. Commit- Chafe, W. L. & Nichols, J. (Eds.) (1986). Evidentiality: and visible social media (Thielmann 2012: ment: Dynamics of Political Hashtags in Russian The Linguistic Encoding of Epistemology. Norwood: Twitter. Retrieved from Ablex. 100) and by the communicative practices http://ssrn.com/abstract=2034506. performed in social media platforms – as

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 70

Dang-Anh, M. & Eble, M. (2013). “Watch out, organ- Gaffney, D. & Puschmann, C. (2014). Data Collec- Krämer, S. (1991). Zur Begründung des Infinitesi- ize, inform yourself!”: Tracing the Dynamics of tion on Twitter. In K. Weller et al. (Eds.), Twitter malkalküls durch Leibniz. Philosophia Naturalis, Twitter Discourse on Anti-Nazi Street Protests. and Society (pp. 55-67). New York: Peter Lang. 28.2, 117-146. Selected Papers of Internet Research, IR 14. Re- Gerbaudo, P. (2012). Tweets and the Streets: Social Linell, P. (1982). The Written Language Bias in Linguis- sistance + Appropriation. Retrieved from Media and Contemporary Activism. London, New tics. Linköping: University of Linköping, Dept. of http://spir.aoir.org/index.php/ York: Pluto Press. Communication Studies. spir/article/view/729. Gillespie, T. (2010). The Politics of 'Platforms'. New Martín Rojo, L. (2014a). Occupy: The Spatial Dy- Dang-Anh, M. et al. (2013a). Kontextualisierung Media & Society, 12.3, pp. 347-364. namics of Discourse in Global Protest Move- durch Hashtags. Die Mediatisierung des politi- Grube, G. (2005). Autooperative Schrift – und eine ments. Journal of Language and Politics, 13.4, pp. schen Sprachgebrauchs im Internet. In H. Diek- Kritik der Hypertexttheorie. In S. Krämer (Ed.), 583-598. mannshenke & T. Niehr (Eds.), Öffentliche Wörter: Schrift: Kulturtechnik zwischen Auge, Hand und Martín Rojo, L. (2014b). Taking over the Square: Analysen zum öffentlich-medialen Sprachgebrauch Maschine (pp. 81-114). München: Fink. The Role of Linguistic Practices in Contesting (pp. 137-159). Stuttgart: ibidem. Günthner, S. & Luckmann, T. (1995). Asymmetries Urban Spaces. Journal of Language and Politics, Dang-Anh, M. et al. (2013b). Die Macht der Algo- of Knowledge in Intercultural Communication: 13.4, pp. 623-652. rithmen – Selektive Distribution in Twitter. In M. The Relevance of Cultural Repertoires of Com- McEnery, T. & Hardie, A. (2012). Corpus Linguistics: Emmer & I. Stapf (Eds.), Authentizität in der communicative Genres. Retrieved from Method, Theory and Practice. Cambridge, New putervermittelten Kommunikation (pp. 74-87). http://kops.uni-konstanz.de/handle/ York: CUP. Weinheim: Juventa. 123456789/3694. Neumayer, C. & Valtysson, B. (2013). Tweet against Daphi, P. et al. (2015). Protestforschung am Limit: Eine Halavais, A. (2014). Structure of Twitter: Social and Nazis? Twitter, Power, and Networked Publics soziologische Annäherung an Pegida. Berlin. Re- Technical. In K. Weller et al. (Eds.), Twitter and in Anti-Fascist Protests. MedieKultur, 29.55, pp. trieved from http://www.wzb.eu/sites/de- Society (pp. 29-41). New York: Peter Lang. 3-20. fault/files/u6/pegida-report_berlin_2015.pdf. Honeycutt, C. & Herring, S. C. (2009). Beyond Mi- NOFRAGIDA. (2015). Facebook page of Dayter, D. (2015). Small Stories and Extended Nar- croblogging: Conversation and Collaboration via NOFRAGIDA. Retrieved from ratives on Twitter. Discourse, Context & Media. Twitter. In R. H. Sprague (Ed.), Proceedings of the https://www.facebook.com/NOFRAGIDA1. Retrieved from http://www.science di- 42nd Annual Hawaiʻi International Conference on NOPEGIDA. (2015). Facebook page of NOPEGIDA. rect.com/science/article/pii/S22116958150001 System Sciences. 5-8 January, 2009, Waikoloa, Big Retrieved from 85. Island, Hawaii (pp. 1-10). Los Alamitos, CA: IEEE https://www.facebook.com/nopegida. Entman, R. M. (1993). Framing: Toward Clarification Computer Society Press. O'Halloran, K. (2010). How to Use Corpus Linguis- of a Fractured Paradigm. Journal of Communica- Klein, J. (2014). Sätze in der Politik – Strukur. Sali- tics in the Study of Media Discourse. In A. tion, 43.4, pp. 51-58. enz. Resonanz. In J. Klein (Ed.), Grundlagen der O'Keeffe & M. McCarthy (Eds.), The Routledge Felder, E. (2012). Pragma-semiotische Textarbeit Politolinguistik: Ausgewählte Aufsätze (pp. 115- Handbook of Corpus Linguistics (pp. 563-577). und der hermeneutische Nutzen von Kor- 126). Berlin: Frank & Timme. London, New York: Routledge. pusanalysen für die linguistische Mediendis- Klemm, M. & Michel, S. (2014). Big Data – Big Prob- O'Keeffe, A. (2012). Media and Discourse Analysis. kursanalyse. In E. Felder et al. (Eds.), Korpu- lems? Zur Kombination qualitativer und quanti- In J. P. Gee & M. Handford (Eds.), The Routledge spragmatik: Thematische Korpora als Basis diskurs- tativer Methoden bei der Erforschung politi- Handbook of Discourse Analysis (pp. 441-454). linguistischer Analysen (pp. 116-174). Berlin / scher Social-Media-Kommunikation. In H. London, New York: Routledge. Boston: de Gruyter. Ortner et al. (Eds.), Datenflut und Informationska- Page, R. E. (2012). The Linguistics of Self-Branding Fuchs, C. (2014). Social Media: A Critical Introduction. näle (pp. 83-98). Innsbruck: Innsbruck University and Micro-Celebrity in Twitter: The Role of Los Angeles et al.: SAGE. Press.

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 71

Hashtags. Discourse & Communication, 6.2, pp. Pomerantz, A. (1984). Agreeing and Disagreeing Thimm, C. et al. (2014). Mediatized Politics – Struc- 181-201. with Assessments: Some Features of Pre- tures and Strategies of Discursive Participation Parkinson, H. J. (2015). Digging into Twitter Tun- ferred/Dispreferred Turn Shapes. In J. M. Atkin- and Online Deliberation on Twitter. In A. Hepp nels: New 'Quote Tweet' Function Launches. Re- son & J. Heritage (Eds.), Structures of Social Ac- & F. Krotz (Eds.), Mediatized Worlds. Culture and trieved from http://www.theguardian.com/ tion. Studies in Conversational Analysis (pp. 57- Society in a Media Age (pp. 253-270). Basing- technology/2015/apr/07/twitter-new-quote- 101). Cambridge: CUP. stoke: Palgrave Macmillan. tweet-function-tunnels. Puschmann, C. & Burgess, J. (2014). The Politics of Tufekci, Z. & Wilson, C. (2012). Social Media and the Paßmann, J. (2014). From Mind to Document and Twitter Data. In K. Weller et al. (Eds.), Twitter Decision to Participate in Political Protest: Ob- Back Again: Zur Reflexivität von Social-Media- and Society (pp. 43-54). New York: Peter Lang. servations From Tahrir Square. Journal of Com- Daten. In R. Reichert (Ed.), Big Data: Analysen Raymond, G. & Heritage, J. (2006). The Epistemics munication, 62.2, pp. 363-379. zum digitalen Wandel von Wissen, Macht und Öko- of Social Relations: Owning Grandchildren. Lan- Wilson, C., & Dunn, A. (2011). Digital Media in the nomie (pp. 259-285). Bielefeld: Transcript. guage in Society, 35.05, pp. 677-705. Egyptian Revolution: Descriptive Analysis from Paßmann, J. (2015). Bewertungssysteme: Medien- Rintel, S. et al. (2013). Knowledge and Asymmetries the Tahrir Data Sets. International Journal of praktiken im Umgang mit Fremden. Zeitschrift in Action: Proceedings of the 2012 Conference Communication, 5, 1248-1272. für Literaturwissenschaft und Linguistik, 45.177, of the Australasian Institute of Ethnomethodol- Wulff, C. (2010). Speech to mark the Twentieth pp. 141-166. ogy and Conversation Analysis. Australian Jour- Anniversary of German Unity. Retrieved from Paßmann, J. & Gerlitz, C. (2014). ‘Good’ platform – nal of Communication, 40.2, pp. 1-8. http://www.bundespraesident.de/SharedDocs/ political reasons for ‘bad’ platform-data: Zur so- Steinberg, R. L. (2014). The Occupy Assembly: Dis- Reden/DE/Christian-Wulff/Uebersetzte Re- zio-technischen Geschichte der Plattformaktivi- cursive Experiments in Direct Democracy. Jour- den/2010/101003-Deutsche-Einheit- täten Fav Retweet und Like. Mediale Kontrolle nal of Language and Politics, 13.4, pp. 702-731. englisch.html;jsessionid=BF9909033012 unter Beobachtung, 3.3. Retrieved from Thielmann, T. (2012). Taking Into Account: Harold DE06F3D564F0C1DAAF9F.2_cid293?nn=2748 http://www.medialekontrolle.de/wp- Garfinkels Beitrag für eine Theorie sozialer Me- 000. content/uploads/2014/09/Passmann-Johannes- dien. Zeitschrift für Medienwissenschaft, 6.1, pp. Zappavigna, M. (2011). Ambient Affiliation: A Lin- Gerlitz-Carolin-2014-03-01.pdf. 85-102. guistic Perspective on Twitter. New Media & So- Penney, J. & Dadas, C. (2014). (Re)Tweeting in the Thimm, C. et al. (2011). Diskurssystem Twitter: ciety, 13.5, pp. 788-806. Service of Protest: Digital Composition and Cir- Semiotische und handlungstheoretische Per- culation in the Occupy Wall Street Movement. spektiven. In M. Anastasiadis & C. Thimm (Eds.), New Media & Society, 16.1, pp. 74-90. Bonner Beiträge zur Medienwissenschaft Vol. 10. Polizei_Ffm. (2015). Twitter Profile of Polizei_Ffm. Social Media: Theorie und Praxis digitaler Sozialität Retrieved from https://twitter.com/Polizei_Ffm. (pp. 265-285). Frankfurt am Main: Peter Lang.

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics 72

Appendix

account /time of # time original text translation referred posting

2015- More than 300 attendants are at the Bereits über 300 Teilnehmer/-innen sind bereits an der #Hauptwache in 1 02-02 Polizei_Ffm #Hauptwache already in #meinFrankfurt #meinFrankfurt #nofragida #fragida 16:50 #nofragida #fragida

2015- Zwischenstand Teilnehmerzah- 02-02 len: #Pegida/ #fragida: 50 17:37 #nofragida: 620 #meinfrankfurt #Hauptwache . Intermediate status attendance numbers: 2 Polizei_Ffm #Pegida/ #fragida: 50 #nofragida: 620 #meinfrankfurt #Hauptwache [photo]

2015- . @Polizei_Ffm The #DankePolizei cannot 3 02-02 @ 17:37 . @Polizei_Ffm Fie #DankePolizei kann nicht zählen: 15 Fragida am Platz ..... count: 15 Fragida on the spot ..... 17:40

2015- @Polizei_Ffm Where are those 50 hiding? 4 02-02 @ 17:37 @Polizei_Ffm Wo haben sich denn die 50 versteckt? #nofragida #nofragida 17:43

2015- .@Polizei_Ffm Are the other 47 in the 5 02-02 @ 17:37 .@Polizei_Ffm Sind die anderen 47 gerade in der Katharinenkirche? Katharinenkirche just now? 17:45

2015- . @Polizei_Ffm Sorry, the #Pegida number was . @Polizei_Ffm Sorry, die #Pegida Zahl war offensichtlich zu hoch. Aktuell haben 6 02-02 Polizei_Ffm too high, obviously. Currently, the colleagues die Kollegen 30 gezählt. Danke für den Hinweis. 17:53 have counted 30. Thanks for the pointer.

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics Mark Dang-Anh & Jan Oliver Rüdiger | From Frequency to Sequence 73

2015- Zwischenstand Teilnehmer- 02-02 zahlen: #Pegida/ #fragida: 18:42 60 #nofragida: 1180 #meinfrankfurt #Hauptwache .

Intermediate status attendance numbers: 7 Polizei_Ffm #Pegida/ #fragida: 60 #nofragida: 1180 #meinfrankfurt #Hauptwache [photo]

2015- Here the official attendance numbers from Hier die offiziellen Teilnehmerzahlen des heutigen Abends: #Pegida/ #Fragida: 8 02-02 Polizei_Ffm tonight: #Pegida/ #Fragida: 85 #nofragida: 85 #nofragida: 1200 #meinfrankfurt #Hauptwache. 19:55 1200 #meinfrankfurt #Hauptwache.

2015- The @Polizei_Ffm has counted: 85 #Pegida- Die @Polizei_Ffm hat durchgezählt: 85 #Pegida-Demonstranten standen 1200 9 02-02 RTL Hessen demonstrants opposed 1200 counter- Gegendemonstranten gegenüber. 20:14 demonstrators.

Table 1: Postings for sequence analysis

10plus1: Living Linguistics | Issue 1 | 2015 | Media Linguistics