From News to Disinformation: Unpacking a Parasitic Discursive Practice of Czech Pro-Kremlin Media

ABSTRACT

This paper examines a “parasitic discourse behavior,” a strategy used by the Czech antisystem media. A working hypothesis about the nature of parasitization was drawn from close reading: the antisystem media, while creating a false impression of mainstream journalism, attempt to reframe events with a set of recurrent associations. This hypothesis was tested with a combination of three quantitative methods: (1) Keyword analysis for identification of prominent topics, which are further analyzed by (2) Companions and (3) Market Basket Analysis (MBA). Companions, a new method comparing occurrences of keywords in time, reveals antisystem’s imitation of the mainstream, and MBA shows its production of distinct associations to frame trendy news items. As a proof of concept, four instances of parasitization were identified during May–August 2020 – topics that attracted public attention for a sufficiently long timespan to come under the radar of our methods: Belarus presidential elections, followed by brutality on the protesters, the police killing of George Floyd in the US, and the subsequent events against systemic racism, the COVID-19 pandemic, and the dispute over the WWII monuments in the Czech Republic. Associations accompanying all these topics weave a set of consistent narratives: anti-West, pro-Kremlin, and a strong anti-Ukrainian stance.

KEYWORDS:

• Keyword Analysis
• Market Basket Analysis
• Anti-system Media
• Framing
• Disinformation
• Czech
• Corpus

1. Introduction

This study examines the “parasitic discourse behavior” of the Czech anti-system online media outlets – how they spawn their own brand of narratives, mimicking current events covered by the mainstream media; i.e., they parasitically “feed on” the mainstream news as a host to disseminate a persistent line of thought. The characteristic properties of anti-system media outlets and their distinct approach to current events vis-à-vis mainstream news-reporting outlets were analyzed with a combination of three quantitative methods: keyword analysis (Scott 2020), Market Basket Analysis (Han et al. 2011), and Companions, a tool introduced for the first time in this paper.

Previous studies focused on limited aspects of anti-system media discourse, exploring the simplest differences between antisystem (ANTS, for more detailed description see sec. 4)² and mainstream (MS) media, e.g., the distinct topics addressed by ANTS with the help of keyword analysis: e.g., ANTS portals in 2018 overused words such as Anglo-Zionist, dollarization, anti-Russian (Fidler and Cvrček 2020). ANTS-specific dehumanizing associations to frame³ migrants and its persistent narratives (e.g., conspiracies and the negative role of the West) were probed with the help of Market Basket Analysis (Cvrček and Fidler 2022).

While Market Basket Analysis (MBA) alone provides a static shot of how events are explained or framed, the current study adds Companions to capture how ANTS engages with particularly trendy news topics over time in relation to MS. As the media discourse in Czech represents part of the public discourse of East European nations whose recent accession to NATO was contested by Russia, it promises to bring valuable insights into the mechanism of hybrid warfare that preceded the Russian invasion of Ukraine.

First we present the working hypothesis based on the qualitative probe (sec. 2 and 3), then we describe the data and explain how the quantitative methods were set up to test the working hypothesis (sec. 4 and 5). Finally, in sec. 6 results and their interpretation are discussed.

2. Working Hypothesis Based on Initial Qualitative Probe

Based on an intuitive manual inspection of ANTS texts from the period of summer 2020 below, we discuss its coverage of two news topics suggesting a certain discursive pattern. ANTS’ major discursive traits can be summarized and posited as a working hypothesis that consists of two parts:

A.	ANTS engages with the trendy news topics raised by MS in a way that resembles the latter. ANTS, in other words, impersonates news reporting media portals.
B.	Simultaneously, ANTS subjects the news topic to conceptual mutation over time, with an aim to spread its own brand of narratives.

This discursive practice can be compared to a parasite behavior. Pretending to be like the mainstream media, ANTS uses a news topic raised by MS as a “host,” using it as a space to multiply its narratives. As a byproduct of presenting such “alternative” narratives, ANTS essentially discredits the mainstream media; like a parasite, ANTS thus spreads its narratives at the expense of MS.

We carried out close reading of manually selected texts that contain two lemmas: Floyd and vakcína ‘vaccine.’

2.1 Case Study 1: Floyd

The first ANTS article on George Floyd is dated May 28, one day after the first appearance in the Czech MS media. Compare the following two titles reporting Floyd’s death:

(1)	MS Title: Nemůžu dýchat, prosil zadržený černoch bělošského policistu. Marně. (05-27) ‘I can’t breathe, the detained black man begged the white police officer. In vain.’
(2)	ANTS Title: OSN vyzvala k ukončení rasového násilí ve Spojených státech. V Minneapolisu se lidé mstí za smrt černocha (05-28) ‘The UN called for an end to racial violence in the United States. In Minneapolis, people take revenge for the death of a black man.’

The titles themselves already point to different stances between ANTS and MS. ANTS implies that the United States has never eradicated racially motivated violence. The use of mstít se ‘revenge’ in ANTS implies vendetta-type protests in response to the killing of Floyd. MS uses zadržený ‘detained,’ highlighting the injustice against Floyd by stating he was detained (but not proven guilty). The adverb ‘in vain’ marks empathy towards the powerless victim.

ANTS later expands narratives that depict the US in disarray, conspiracies of the elites, mainstream media as spreading fake news (Floyd’s death as a hoax, disintegration of the United States, the “unfair” glorification of Floyd by educators, corporate America, mainstream media, and the “elites” of the US):

(3)

ANTS Title: Dr. Heartstrong: George Floyd je naživu. Celá situace byla zmanipulovaná (06-09) ‘Dr. Heartstrong: George Floyd is alive. The whole situation was rigged’

(4)

ANTS: V závěru svého dopisu profesor kri[ti]zuje vedoucí činitele univerzity a zvlášť její katedru historie za velebení George Floyda, zpustlíka a na drogách závislého násilníka, který ubližoval ženám a zvlášť pak černým ženám. Avšak společným úsilím Univerzity Kalifornie v Berkeley, korporativní Ameriky, většinových mainstreamových médií a vůbec současných elit USA se Floyd stal kulturním hrdinou položeným do zlaté rakve (06-29) ‘In the conclusion of his letter, the professor criticizes the leaders of the university, and especially its history department, for glorifying George Floyd, a thug and drug-addict rapist who harmed women, and especially black women. But with concerted efforts of the University of California at Berkeley, the corporate America, the majority mainstream media, and in general the current US elites, Floyd has become a cultural hero laid in a golden coffin’

2.2 Case Study 2: Vaccine

The word vaccine starts appearing in early June of 2020 when the Covid-19 vaccines were being developed. It is perhaps one of the most fertile stories for ANTS, which spawns multiple narratives. ANTS’ first articles containing vakcína ‘vaccine’ could be considered to mimic a news reporting style; even the timing is the same as in MS.

(5)	MS Title: Vakcína proti Covidu-19 bude podle AIFP počátkem příštího roku (06-04-2020) ‘Vaccine against Covid-19 will be [available] early next year, according to AIFP [The Association of Innovative Pharmaceutical Industry in Czechia].’
(6)	ANTS Title: Jak to vypadá s vývojem vakcíny proti Covidu-19? Podle výkonného ředitele AIFP by mohla být dostupná začátkem roku 2021 (06-04-2020) ‘What about the development of a Covid-19 vaccine? According to the AIFP CEO, it could be available in early 2021.’

One day later, however, ANTS suggests victimization of Czechs by the Americans.

(7)

ANTS: Český stát bude asi nakupovat vakcíny vyrobené v České republice na závodě s dlouhou tradicí sahající ke vzniku Československa a bude za to platit Američanům. To je k vzteku. (06-05-2020) ‘The Czech state will probably shop vaccines manufactured in the Czech Republic at a plant with a long tradition dating back to the foundation of Czechoslovakia and will pay the Americans for it. This is infuriating.’

Then articles related to vaccination as issues on the international scale appear, e.g., vaccination as a global conspiracy.

(8)

ANTS Title: Good Club: Nejmocnější spiknutí dějin? Gates již ovládá světovou farmacii. Očkováním proti Covid-19 k planetárnímu otroctví. Soros nechybí. (06-22-2020) ‘The Good Club: History’s Most Powerful Conspiracy? Gates already controls the world’s pharmacy. By vaccinating Covid-19 for planetary slavery. Soros is not missing.’

In January and February 2021 ANTS presents the “damaging effects” of vaccine to health and about the “globalists” jeopardizing individual freedom.

(9)

ANTS: Očkování vakcíny a virová hrozba tak poslouží jako záminka k vytvoření státních databází očkovaných obyvatel a k zavedení propustek v podobě covidových pasů. Tím získají globalisté totální kontrolu nad populací (01-15-2021) ‘Vaccination and the viral threat will thus serve as a pretext for creating state databases of vaccinated residents and for implementing passes in the form of Covid passports. This will give the globalists total control over the population.’

Covid-related texts also contain a narrative of migration crisis to Europe, which allegedly is said to lead to a war.

(10)

ANTS: Není to tak, že máme imigrační krizi za sebou, máme ji před sebou. To, co jsme viděli dosud, to byla jenom ochutnávka. Bohužel může dojít k vyprovokování války v Evropě (02-27-2021) ‘It’s not that we have the immigration crisis behind us, we have it ahead of us. What we’ve seen so far, it’s just a tasting-sample. Unfortunately, it may lead to provocation of war in Europe’

3. Pilot Probes

Pilot quantitative probes are also consistent with the qualitative observations in Section 2. In Figure 1 the frequency analysis shows how the topic appears in the mainstream media and is then taken up by ANTS after a short delay. The blue line represents the frequencies of Floyd in MS, the tomato line in the ANTS, showing some delay.

Figure 1. Frequency development of the word Floyd in ANTS and MS during May 1 – August 31, 2020.

As the case continues to unfold, ANTS “grabs” the topic and addresses it extensively, possibly adding ANTS-specific associations.

Likewise, ANTS parasitizes the topic of vaccines. The graph in Figure 2 shows a result of our pilot study using MBA (cf. Section 5). The bars represent the number of words associated with vaccine in three periods of time and how these concepts are shared between the media types. It illustrates how ANTS first sponges off MS associations but frames the topic of vaccination with an increasing number of its own associations as time goes by, to create its own narratives (Cvrček and Fidler 2021).

The chart shows not only that the association array (cf. section 5.3) increased over time, but also that the MS associations in one period become shared associations (they are absorbed by ANTS) while ANTS adds its own new associations in each period (Cvrček and Fidler 2021).

Figure 2. The range of associations with vaccine in three time periods (from June 2020 to February 2021).

4. Data

The working hypothesis was tested on the data of the ONLINE corpus (Cvrček and Procházka 2020), which contains texts from Czech internet news websites (approx. 4 mil. tokens a day). The classification of these websites (including MS and ANTS) stems from the audience-based typology created by Šlerka (2018). It groups portals into clusters in terms of the similarity of their readers, more precisely, their online behavioral patterns, such as visits to web sites (based on Alexa Rank https://www.alexa.com/) and sharing and liking social media articles (based on the CrowdTangle service). The clustering based on audience overlap does not stem from extracting linguistic characteristics, topic preferences, or political stance of the media portals. The only part of the classification subject to researcher interpretation is the labeling of each cluster; these are derived from the features of the site that can be considered as a cluster prototype based on its size and position.

In this study we used ONLINE corpus data from May 1 to August 31, 2020. This four-month period was chosen because of the diversity of topics covered by news media (cf. Section 6.1). Since there were more sources in the mainstream part of the corpus than we would have needed in analysis, we used texts from only the largest and most influential MS portals in analysis. The overall numbers for the data used in the analysis are summarized in Table 1.

Table 1. Size of Corpus Used

	Tokens (mil.)	Articles
Anti-system (ANTS)^4	9.3	4,605
Mainstream (MS)^5	19.0	14,099

5. Methods

Based on our qualitative observations (cf. sec. 2) we could speculate that parasitic ANTS behavior is a complex phenomenon that can be operationalized as follows:

i.	ANTS and MS share prominent concepts (ANTS engages with MS trending topics)
ii.	ANTS and MS share a similar frequency development showing association between MS and ANTS over time (with possible lag in ANTS)
iii.	ANTS and MS show conceptual changes in the association array (cf. section 5.3 below) over time

(i) and (ii) indicate the working hypothesis Part A (ANTS impersonating MS), and (iii) the working hypothesis Part B (ANTS using trendy news topics to spread its own narratives).

To examine these components with the least degree of subjectivity and to test whether the phenomenon of parasitic behavior is widespread and not anecdotal (cf. the discussion about synergies between qualitative and quantitative approaches, e.g. by Baker et al. 2008 or Gabrielatos and Duguid 2015), we applied a combination of three corpus-linguistic methods to the data: Keyword analysis (KWA), Market Basket Analysis (MBA) and Companions.

5.1 Keyword Analysis

Keyword Analysis (KWA, cf. Scott & Tribble 2006) compares the frequencies of lemmas in the target text(s) with those in the reference corpus and extracts words (keywords, KWs) that appear in the target text(s) more frequently than expected against the background of a reference corpus. The extracted KWs refer to concepts, entities, and events that are likely viewed as striking (we use the term “topic” and “keywords” interchangeably in describing these prominent words) by readers who are exposed to the pattern of language use represented by the reference corpus. This study uses ANTS and MS internet newspaper articles as target texts (each article yields one KW list), and offline journalistic texts from 2015–2018 containing traditional Czech tabloids and broadsheets as the reference corpus (i.e., part of the SYNv8 corpus, Křen et al. 2019).⁶

In total, KWA yielded more than 140 thousand different KWs in ANTS and 350 thousand KWs in MS (many of them repeated in more than one text). In order to follow how ANTS news topics wax and wane, we selected the KWs that have a reasonable longevity over time. Given the length of the observed period (4 months), we chose a minimum of 20 days as the “long-term KW” prominence threshold. 932 KWs met this criterion.

5.2 Companions

A new tool, Companions, measures the extent to which ANTS mimics MS in covering a news topic over time; it tracks KW frequency development. If one media class shows similar rises and falls in frequency simultaneously or slightly later than the other, the former is presumably mimicking the latter in its discourse (cf. Figure 1). Synchronization strength can be measured by cross-correlation (used in signal processing, for example), which compares frequency development curves of two words (or one word in two media segments) over time and evaluates the similarity of their shapes. In addition to the correlation coefficient (r), it also calculates the lag between curves, by which one media segment is delayed in its development relative to another.

Companions was originally developed to measure how two words, triggered by some real-life event, start occurring in the same time slots; in this sense the two words become “companions” in discourse. The study of similar coinciding frequency developments has an enormous potential for discourse analyses, e.g., the breakout of Covid-19 gave rise to the metaphor of “war against Covid,” where two words – Covid-19, the disease, and the originally warfare-related term frontline – started to appear together as “companions.” The current study uses this tool to test whether two media classes behave like companions with respect to the frequency of one and the same word. The tool examines the synchronized appearance of a word and the degree to which its growth and decline show a similar pattern in ANTS and MS.

Companions was applied to all 932 previously-identified long term KWs. This allowed us to zoom in on 46 KWs that show significant and at least moderate correlation (r > 0.5). We found that 40 of them to varying degrees relate to news topics that emerged during the four-month period (cf. section 6.1), the remaining 6 KWs refer to name of the months: frequent time expressions are characteristic of periodicals but do not relate to specific news events.

5.3 Market Basket Analysis

The results from the previous two procedures were fed into Market Basket Analysis (MBA). MBA indicates the intensity and the distinct way news topics are framed in ANTS in contrast to MS. It is a data-mining method originally developed in marketing; its purpose was to identify the relationships between items in a shopping cart. The method sifts through many transactions and uses probabilistic methods to determine the connection strength between the purchase of item X and other items (A, B, C). When applied to texts (Cvrček and Fidler 2022), MBA helps determine the extent to which a long-term topic X is likely to co-occur with other topics A, B, C in the same text (cf. Cvrček and Fidler 2021, 2022). When the co-occurrence is determined to be strong and regular⁷, A, B, and C are considered as forming X’s “association array” – a frame into which X is regularly embedded in texts. The KWs (A, B, C), the content of the association array (henceforth AA) for the word X, will be referred to as “associated KWs” of the keyword X.

AA differs from collocation of KWs or KW-links which are sometimes used to harvest associations. While the latter methods concern how the immediate context imbues the semantics of the KW, AA informs of what concepts appear together within the entire text (the whole article), regardless of the distance between KWs. Roughly speaking, collocations and KW-links suggest the understanding of a word (e.g., the tendency of the word immigrant to be understood as illegal because of the frequency proximity between the two words), while MBA suggests the likely conceptual framing of a topic (e.g., immigrants are sent to Europe by globalists and NGOs to destroy the European identity).

Since we wanted to be able to capture the trends in the data, we divided the period into two parts (May 1 – June 30 and July 1 – August 31) and compared them. For each of the aforementioned 40 KWs, we created 4 association arrays (2 for each media type). AAs that are found in May-June are recorded as AA.ANTS.pre and AA.MS.pre for ANTS and MS respectively; those found in July-August as AA.ANTS.post and AA.MS.post for ANTS and MS (cf. Appendix). Upon close inspection of these KWs, we focused on a group of eleven KWs, each of which is most clearly tied to a single specific news topic (Table 3) and can be used as a typical representative of the topic. This down-sampling of data was necessary to clearly see the amount and the types of distinct associations consistently found in ANTS in contrast to MS.

The relationship between each method and its purpose(s) is summarized in Table 2.

Table 2. Methods and Purposes

Method	Purpose(s)
Keyword Analysis	Identification of KWs that have longevity of at least 20 days
Companions	Testing working hypothesis Part A: ANTS general tendency to mimic MS
MBA	Testing working hypothesis Part B: changes in conceptual framing of news topics over time in ANTS

6. Results

This section presents the results with an aim to test the working hypothesis formulated in Section 2. We first examined all 46 long-term KWs. Then we selected the most relevant KWs in the end for the purposes of testing the working hypothesis.

6.1. Working Hypothesis Part A: ANTS Mimicking MS

As mentioned above, Companions identified 40 KWs that show at least moderate correlation (r > 0.5), i.e. the KWs that showed similar frequency fluctuations in both MS and ANTS. These KWs relate to the following news topics to varying degrees (the relationship between news topics and the KWs was assessed by inspecting the texts in which they occurred):

a.	the Belarus presidential elections, followed by police brutality on the protesters
b.	the police killing of George Floyd, and the subsequent events against systemic racism
c.	the Beirut explosion
d.	summit meetings (discussions on economic sanctions, and the EU recovery fund, security)
e.	Covid-19
f.	the economic crisis (partly caused by Covid-19)
g.	the controversy over WWII monuments: the removal of the statue of General Konev as the WWII Red Army liberator of Czechoslovakia; a new monument honoring General Vlasov; the alleged assassination plan on the local Czech politicians by the Russian secret service; and the expulsion of Russian diplomats.
h.	Czech domestic politics

The Working hypothesis Part A was tested using Companions, specifically by its cross-correlation coefficient of frequency development. This represents the degree of similarity for each word as it appears in ANTS and MS. The lag scores for words exhibiting moderate or strong correlation (r > 0.5) are summarized in percentages in the chart below. For the actual lag scores, see the first column in Appendix.

The relevant lag values for analysis of topics in media range from 0 (no lag) to 3 or 4 days. This corresponds to the time which is needed for ANTS to adopt a new story first released in MS, while the topic is still trendy. In fact, nearly 80% of the 40 KWs show lag scores of 0 to 2 and exhibit similar rises and falls in frequency. ANTS engages with a news topic in a way similar to MS: when MS is more engaged with a news topic, so is ANTS. The lag scores support our hypothesis that ANTS is more likely to be imitating MS, rather than the other way around.

Even more informative are the cross-correlation of KWs that are most reliably tied to a single specific topic. These are KWs tied to four topics referred to as: Belarus-KWs, BLM-KWs, Covid-KWs, and Liberation-KWs named after the news topics (Table 3). The Belarus-KWs (Lukashenka, Belarus, Belarusian, Minsk) yielded very high scores (cross-correlation above 0.76). The BLM-KWs (Floyd, racism) also scored high (0.77 and 0.62). The Covid-KWs (coronavirus, epidemic, pandemic) and the Liberation-KWs (red, Czechoslovakia) also yielded relatively high scores above 0.60.

6.2 Working Hypothesis Part B: Conceptual Mutation over Time

Keyword analysis and Companions might indicate how ANTS and MS share prominent news topics and how ANTS mimics MS in terms of frequency, but they do not let us know to what extent ANTS frames the same news topics in the same manner as MS. This section examines the association arrays (AA) extracted by MBA. The size of an association array (i.e., the number of associated words) in the first and the second half of the four-month period shows whether ANTS frames a news topic at varying intensities (cf. the growing number of associations to vaccine in Figure 2). Differences in the size of AA between ANTS and MS suggest different types of framing the same news topic. The associated KWs in an AA point to the way in which ANTS weaves its narratives around a news topic. We will focus on the 11 KWs that are unambiguously tied to specific and trendy news topics listed in the previous section (Table 3).

Figure 3. Lag scores of the 40 ANTS KWs in relation to MS

Table 3. KWs tied to most trendy news topics

	The Czech original KWs	translation
Belarus-KWs	Lukašenko, Bělorusko, běloruský, Minsk	Lukashenka, Belarus, Belarusian, Minsk
BLM-KWs	Floyd, rasismus	Floyd, racism
Covid-KWs	koronavirus, epidemie, pandemie	coronavirus, epidemic, pandemic
Liberation-KWs	rudý, Československo	red^8, Czechoslovakia

Before discussing the results, there are two important points about the nature of news topics. First, it is necessary to consider that news topics develop in different directions. The Belarus protests against the presidential elections intensified and the police oppression of the demonstrators became increasingly brutal. The police murder of George Floyd developed in multiple directions under a large theme of systemic racism. It sparked demonstrations, riots, police reforms, and dismantling of historical monuments within and beyond the United States. The news on COVID-19 waned during the four-month period as the spread of the illness temporarily declined in the Czech Republic in the summer of 2020. Discussion on the liberation of Czechoslovakia also died down, as the assassination attempt on the two Czech politicians who challenged the Russian interpretation of the Red Army’s liberation of Czechoslovakia did not take place⁹. The size of association arrays is expected to interact with these real-life fluctuations. Second, it is expected that the news coverage of Belarus protests and Covid-19, by virtue of being geopolitical-international and current, would attract more public attention than the coverage of BLM (which takes place largely across the ocean) and the liberation of Czechoslovakia (which took place in the past).

Figure 4 summarizes the size of AAs in ANTS and MS in the two periods. The red bars show ANTS-AAs and the blue ones MS-AAs. The KWs, presented in the x-axis, are arranged in groups to reflect the four news topics. The y-axis represents the size of AAs (the number of associated KWs).

Figure 4. Size of AA for KWs representing topic subjected to parasitization.

What is immediately noticeable is the larger size of ANTS-AAs compared to MS-AAs which means that ANTS creates larger network of associations for each topic.

The second noticeable characteristic of ANTS is that it is overall more sensitive to public attention than MS. This is visible in the ANTS-AAs for the Belarus protests and COVID-19. ANTS creates much more associations than MS as the Belarus situation intensifies. The abrupt hike in the size of ANTS-AA suggests attempts to target a news topic that is increasingly attracting public attention. Conversely, the association arrays of the COVID-KWs contract more drastically in ANTS than in MS. In view of the declining numbers of the COVID patients during this period, the steep contraction in ANTS can be seen as the mirror image of the Belarus-KW ANTS-AAs; ANTS creates a wider range of associations than MS only while the public is keenly interested in COVID-19.

Compared to these two news topics, the picture is not as dramatic with the AAs for the other two topics: BLM and liberation of Czechoslovakia. Nonetheless, it is noteworthy that ANTS is more sensitive to the consequences of George Floyd’s death, showing a spike in the size of AA for racism during the second half of the four-month period. As for the liberation of Czechoslovakia, ANTS’ AAs show persistent associations (unlike MS that has null associative KWs) and abrupt decline in the size of AAs as the news topic waned.

ANTS creates more associations when the news topic attracts public attention. It thus resembles a parasite that proliferates extensively when it finds the best fertile host. Conversely, when the gravity of the news topic wanes, ANTS abandons it abruptly, a behavior akin to a parasite’s leaving a dying host that is no longer useful for reproduction.

In order to examine the nature of ANTS-specific narratives, however, we must look more closely at the associated KWs for the four news topics. We first established the following inventory of associated KWs that recur in ANTS after going through associated KWs in both ANTS and MS.

Table 4: KWs often encountered in ANTS texts

	Associated KWs	English translation
References to the West	USA, Amerika, americký, Trump, EU, Brusel, NATO, západ, západní	USA, America, American (adj), Trump, EU, Brussel, NATO, west, western
References to Russia	Rusko, ruský, Rus, Putin, Moskva, Kreml, sovětský, SSSR	Russia, Russian (adj), Russian (noun), Putin, Moscow, Kremlin, Soviet (adj), USSR
References to threats: military and security-related forces mass violence that could provoke mobilization of military forces	armáda, vojenský, voják, tajný, vnitro, nepokoj, násilí, nenávist	army, military (adj), soldier, secret [services], internal [affairs], unrest, violence, hatred
References to threats: death, illness, acute human needs	úmrtí, smrt, mrtvý, nemoc, zemřít, onemocnění	fatality, death, dead, illness, to die, disease
References to Ukraine	Ukrajina, Kyjev, Donbas, Doněck, doněcký, Zelenský, Luhansk, luhanský	Ukraine, Kyiv, Donbass, Donetsk(noun), Donetsk (adj), Zelensky, Luhansk, Luhansk (adj)
Other unexpected KWs not found in the MS AAs	Bill, Gates, Majdan, bílý, běloch	Bill, Gates, Maidan, white (adj), white person (noun)

The presence of the associated KWs was noted in Table 5. The notation ANTS in the table indicates that at least one associated KW in the group was found exclusively in ANTS; ANTS + MS indicates that the associated KW was found in both MS and ANTS.

Table 5: Groups of associated KWs found in ANTS AAs

News topic	KWs	The West	Russia	Threats		Ukraine	other notable associated KWs
				military and security-related forces and mass violence that could provoke mobilization of military forces	death, illness, acute human needs
Belarus presidential elections and demonstrations	Lukashenka	ANTS	ANTS + MS	ANTS		ANTS	Maidan
	Belarus	ANTS	ANTS + MS	ANTS		ANTS	Maidan
	Belarusian	ANTS	ANTS + MS	ANTS		ANTS	Maidan
	Minsk	ANTS	ANTS + MS	ANTS		ANTS
COVID-19	coronavirus	ANTS + MS	ANTS + MS	ANTS	ANTS + MS	ANTS	Bill, Gates
	epidemic	ANTS	ANTS	ANTS	ANTS + MS	ANTS	Bill, Gates
	pandemic	ANTS + MS	ANTS	ANTS	ANTS + MS	ANTS
BLM	Floyd	ANTS + MS		ANTS	ANTS + MS		white(adj)
	racism	ANTS		ANTS			white person(noun)
Dispute over liberation of Czechoslovakia	red	ANTS	ANTS	ANTS
	Czechoslovakia	ANTS	ANTS	ANTS

At first sight, the table shows ANTS’ strong interest in threats associated with military and security. It is noteworthy that such associations consistently appear even with otherwise health-related topics such as COVID-19. In the remainder of this section, we first generally comment on individual news topics and their associated KWs. We then proceed to a more detailed discussion of illustrative examples.

In conjunction with racism, it is not surprising that the associated KW USA co-occurs with the KW in ANTS. Unlike MS, however, ANTS places unusually persistent emphasis on the US and on hatred (rather than, e.g., the pursuit of racial equality) and unrest (requiring armed responses) in texts about racism (i.e., these words do not enter into the MS-AA); these associated KWs suggest that ANTS narrative might focus much more on the negative consequences of racism permeating the United States than MS. It is also noteworthy that ANTS, unlike MS, yields the associated KW a white man and white.

Likewise, the KW red (Army) is expected to co-occur with references to the USSR and armed threats; this is indeed the case in ANTS, but its use in ANTS is much more persistent than MS, as the latter does not yield any such associated KWs. This contrast again points to ANTS’ unusual and systematic focus on geopolitics where the West and Russia are depicted as the main social actors in potential conflict.

As for the Belarus-KWs, they co-occur with references not only to the West, Russia, and warfare (as expected), but also to Ukraine and Maidan. The associated KWs in MS overlap with the ANTS counterpart only in terms of Russia-related terms. The types of associated KWs then point to a distinct narrative by ANTS that connects the Belarus protests to the West and Ukraine; references to Maidan suggest an analogy drawn between the Belarus protests and Ukraine’s Orange revolution, and points to the potential Ukrainian influence on the Belarus protests, cf. Russia’s growing concern over the Ukrainian Maidan (Kurfürst 2021, 42).

Finally, the COVID-KWs are expected to include KWs referring to health threats. The ANTS AAs, however, also contain KWs that refer to the West and Russia, and most interestingly to Bill Gates. The recurrent KWs Bill and Gates suggest the ANTS-specific conspiracy-based framing of COVID-19 where Bill Gates plays a major role. ANTS’ heavy geopolitical focus in COVID-related texts is different from MS, which is more engaged in presenting data regarding the spread of COVID-19.

The associated KWs extracted from MBA indicate that ANTS differs from MS in excessively focusing on geopolitics where the West and Ukraine play prominent roles. For information beyond this point we must look at the texts that contain these associated KWs. Below are some excerpts from ANTS texts.

Bělorusko ‘Belarus’ occurs in texts that express concerns about NATO’s military involvement. Ukrainians are depicted negatively as Nazis. Words in bold style represents a KW; if underlined, the word is a KW in the second column in Table 5.

(11)

ANTS: NATO oslovilo Bělorusko: Pozorně sledujeme situaci ve vaší zemi. Bělorusko uspořádá vojenské cvičení na hranicích s Litvou. Během cvičení části běloruské armády bude uskutečněn naplánovaný nácvik v bojové přípravě včetně bojových střeleb. (08-19) ‘NATO addressed Belarus: We are closely following the situation in your country. Belarus is organizing a military exercise on the borders with Lithuania. During the exercises by part of the Belarusian army, a scheduled training in combat preparation including combat shooting will be carried out.’

(12)

ANTS: Ukrajinská varianta v Bělorusku nehrozí. Jdi pryč, ty oškliváku! Ukrajinští nacisté nabízejí pomoc běloruským opozičníkům. (08-19) ‘The Ukrainian option does not threaten Belarus. Go away, you ugly man! The Ukrainian Nazis are offering assistance to the Belarusian opposition.’

Below is a sample of the KW racism where protests against racism in the United States are negatively portrayed.

(13)

ANTS Title: BLM a Antifa jsou morbidní rasisté, vyznávající komunismus ‘BLM and Antifa are morbid racists, espousing communism’ ANTS Text: Proto strhávají sochy amerických politiků a hrdinů, kteří se nějak zasloužili o dobro národa, proto žádají nesmyslné náhrady za otroctví, které nikdy nezažili, volají po smrti pro všecky bílé - z toho všeho čiší pouze nenávist, závist a rasismus. (08-10) ‘That’s why they are tearing down statues of American politicians and heroes who have somehow contributed to the good of the nation, that is why they are asking for meaningless compensation for slavery that they never experienced, [that’s why] they are calling for death for all whites - all of this is just hate, envy and racism.’

The KW rudý ‘red’ appears in a story where the soviet heroes of WWII are being denigrated. Russia is represented as an unfairly treated victim:

(14)

ANTS: Jeho generálům a maršálům v čele Rudé armády se stále více dařilo hnát Němce zpět, odkud přišli. […] Na hrdiny, kteří se zasloužili o vítězství nad Hitlerem a na zmaření rozvratných plánů na zničení SSSR se hází špína, zatímco kolaborantům s velkokapitálem se kol hlavy nasazuje svatozář čistokrevných demokratů. (05-04) ‘His generals and marshals at the head of the Red Army were increasingly successful in driving the Germans back from where they had come. […] The heroes who were instrumental in the victory over Hitler and in thwarting the divisive plans to destroy the USSR are being smeared, while the collaborators with big capital are being crowned with the halo of purebred democrats.’

In contrast to Russia, the West (“especially the Anglo-Saxons”) was allegedly planning a massacre at the end of WWII.

(15)

ANTS: V ohnivých smrštích měly zahynout miliony Rusů, stejně jako byl zničen Hamburk, Drážďany, Tokio, … To se chystali učinit se svými “spojenci”. Tedy obvyklý obraz: nejodpornější zrada, krajní podlost a zvěrská krutost - to je vizitka západní civilizace a zvláště pak Anglosasů, kteří zabili nejvíce lidí v celé lidské historii. Nicméně 29. června 1945, den před plánovaným začátkem nové války, Rudá armáda nečekaně změnila svoji dislokaci. (06-11) ‘Millions of Russians were to perish in the firestorms, just as Hamburg, Dresden, Tokyo had been destroyed . . . This was what they were going to do with their “allies”. So, the usual picture: the most heinous betrayal, the most wicked and the most savage cruelty - this is the calling card of Western civilization and especially the Anglo-Saxons, who killed the most people in all of human history. However, on June 29, 1945, the day before the planned start of a new war, the Red Army unexpectedly changed its dislocation.’

ANTS’s text on pandemics and coronavirus weaves a narrative about Bill Gates’ alleged conspiracy:

(16)	ANTS: Celý svět podezřívá Billa Gatese, že tuto globální pandemii způsobil za tím účelem, aby nás mohl očkovat a čipovat. (05-23) ‘The whole world suspects Bill Gates caused this global pandemic with the goal to inoculate and chip us.’

In contrast, in texts on COVID, Russia is – unlike Ukraine – presented as a successful country that the United States envies.

(17)

ANTS: V USA vznikla hysterie kvůli oznámení o ruské vakcíně proti koronaviru. Od okamžiku, kdy do kosmu letěl první Rus a nikoli Američan, nedokáží v USA skousnout jakýkoli úspěch Ruska a snaží se jej všemožně zpochybňovat a popírat. (08-15) ‘In the US, hysteria emerged over the announcement of a Russian vaccine against the coronavirus. From the moment when the first Russian, and not an American, flew into space, the US has been unable to accept any success of Russia, and has tried its best to doubt and deny it.’

(18)

ANTS: Koronavirus na Ukrajině: Pětina nakažených jsou lékaři. Stovky zdravotnických pracovníků absolvují léčení v nemocnicích. (05-02) ‘Coronavirus in Ukraine: A fifth of those infected are doctors. Hundreds of health workers undergo treatment in hospitals.’

7. Conclusions

The aim of this study was to uncover and describe “parasitic” discursive behavior used by the Czech antisystem media. Our working hypothesis based on close reading of text samples was that such discursive “parasitization” consists of two components: mimicking the “host” media (the mainstream) and conceptually mutating news topics to multiply ANTS-specific narratives. In other words, ANTS builds an image of a news portal by reporting on multiple trendy topics (alongside the mainstream media), but it also attempts to reframe events and their actors by connecting them to its own specific associations and to promote own narrative lines.

A combination of three quantitative methods – Keyword Analysis, Companions, and Market Basket Analysis – was applied to data from the Czech antisystem and the mainstream media published in May-August 2020. These methods provided solid empirical evidence for ANTS’ parasitic behavior when covering four major news stories – the Belarus elections, the BLM movement, Covid-19, and discussions about the WWII liberation of Czechoslovakia. These are topics that attracted public attention for a sufficiently long timespan to come under the radar of our methods.

Companions, with its cross-correlation and lag scores, revealed how ANTS imitates MS: the former tends to trace (with a slight delay) the frequency patterns of keywords linked to the news stories. Second, conceptual mutation of news items was confirmed via MBA. In general, the association arrays for the news stories were considerably larger in ANTS than in MS, indicating unexpected framing of news stories. The abrupt hike in the size of ANTS’ associative arrays was observed when a news item drew public attention over time; it shows how ANTS links a trendy topic to a multitude of ANTS-specific associations, thereby creating narratives not found in MS. Conversely, abrupt decline in the size of ANTS’ associative arrays was also found when a news item waned over time; ANTS evidently abandons less newsworthy topics as inadequately fertile (or dying) hosts to multiply its narratives. The text samples containing KWs from ANTS associative arrays served as the most typical examples that best illustrate the major characteristics of ANTS narratives. These texts demonstrated ANTS’ anti-West, pro-Kremlin, and strong anti-Ukrainian stance (“Ukrainians as Nazis”). Whereas the former two were anticipated based on the observations from the initial qualitative probe, the latter was a noteworthy finding obtained with the help of MBA and could be seen as an attempt to prepare the groundwork to justify an imminent drastic action by Russia against Ukraine.

The present paper took the working hypothesis as a starting point. Each of the quantitative methods was designed to test a specific aspect of what might constitute parasitization of mainstream media (cf. Table 2):

• a set of KWs under examination was delimited with respect to their longevity and their reliable connection to specific news items

• the thresholds for cross-correlation and lag were strictly set

• the ANTS association arrays for the relevant KWs were examined both over time and in contrast to the MS association arrays

In short, the existence of parasitization was examined systematically and under strictly set parameters. Besides attesting the parasitic discourse by ANTS, the three methods helped us see what constitutes ANTS narratives. This study revealed the pervasive characteristics of ANTS narratives. It therefore not only strengthened the working hypothesis based on the qualitative probe, but also went beyond observation of selected texts.

It is possible that such recurrent narratives enter the reader’s world knowledge, which in turn becomes easily accessible when interpreting newly arising events, e.g., upon hearing about a crime, readers of anti-immigration texts might be more prone to a false conclusion that it was most probably committed by immigrants (i.e., such knowledge will become part of “frame” in the sense of Tannen 1993; cf. the effect of repeated framing in Lakoff 2004). Use of what the addressee considers most relevant (accessible) to interpret utterances has been shown in the Relevance Theory (Sperber and Wilson 1986/1995/2001). Confirmation of the influence of ANTS discourse, however, is beyond the scope of this paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

2 Czech anti-system (ANTS) portals can be characterized roughly by the following features: (1) The majority of ANTS media outlets maintain an opaque ownership structure, making it challenging to identify the individuals or entities behind them (the only exception is Sputnik News, which openly acknowledges its affiliation with the official Russian press agency). Several ANTS portals, such as Česko aktuálně, reveal domain ownership information but remain silent about the publisher’s identity. Moreover, there are instances where multiple portals appear under the same ownership (e.g., ePortál and EUportál). Some portals registered under company names, like New World Order opposition, do not disclose media ownership data. Aeronet is associated with a company located in the Trump Tower in New York City, but this company is not registered in the state of New York. (2) Anonymous authors and editorial team. (3) ANTS often lack proper citation of sources or provide sources that are either irrelevant or questionable. (4) ANTS are known for disseminating disinformation and pro-Kremlin propaganda. They promote various conspiracy theories, including anti-Islam theories, the “Deep State Theory” alleging the existence of a covert government within a government, the “Theory of New World Order,” and disinformation tailored to the Czech Republic. Additionally, they propagate claims of covert geopolitical operations conducted under a false flag. A detailed description of recurring themes in ANTS portals is provided by the Czech Independent Journalism Foundation https://www.nfnz.cz/dezinformacni-a-konspiracni-media/ (accessed Aug 28, 2023). The recurring ideological undercurrent in ANTS using MBA is examined in Cvrček and Fidler (2022).

3 The term frame is used differently from Tannen 1993, which refers to speech participants’ existing expectations. In this paper the term is used to refer to a set of associations repeatedly linked to a topic, presumably with an intention to later create expectations (Tannen’s frame), which would become part of readers’ knowledge schema.

4 The ANTS servers used for this study are portals classified by Šlerka as political tabloids (parlamentnilisty.cz, prvnizpravy.cz, irucz.ru, zastavmezlodeje.com, necenzurujeme.cz, global.cz, ireporter.cz) and almost 40 portals classified as antisystem servers (the majority of articles come from sputniknews.cz, pravdive.eu, novarepublika.cz, pravyprostor.cz, infokuryr.cz, ceskoaktualne.cz, nwoo.org, protiproud.cz, czechfreepress.cz, svobodnenoviny.eu, euportal.cz, aeronet.cz, ac24.cz, casopis-sifra.cz, zvedavec.org, eportal.cz). Note that some of the URLs might not be accessible as the Czech authorities in February 2022 decided to shut down some of the pro-Russian web portals after the start of Russian aggression on Ukraine. The texts are still available in the ONLINE corpus.

5 MS servers: seznamzpravy.cz, irozhlas.cz, novinky.cz, idnes.cz, denik.cz, lidovky.cz, ceskenoviny.cz, impuls.cz, reflex.cz, respect.cz, ihned.cz, nova.tn.

6 We are aware of a number of differences between online and offline journalism. However, the exact composition of this reference corpus used is hardly relevant to the resulting analysis. The primary aim of the analysis is to compare MS and ANTS, the corpus of offline journalistic texts serves only as a reference point to which we compare the texts from both media classes. The choice of the reference corpus was driven by pragmatic concerns: the two main segments of online journalism (MS and ANTS) are the subject of the analysis, so we do not have online journalistic texts that would be usable for the reference corpus. We therefore used data from offline journalism, which is the most similar to the texts under study in terms of genre-register characteristics and which represent a general usage of written contemporary Czech.

7 As “strong and regular” are considered those associations that co-occur with the seed word with a certain frequency and dispersion across texts (i.e., it is not a co-occurrence appearing only in one or a few texts). MBA can quantify these features using support, confidence and lift indices. (Cvrček and Fidler 2022)

8 This term is associated with communism: e.g., Rudá armáda ‘The Red Army’ and Rudé právo ‘The Red Justice/the Red Right’ (the newspaper of the Communist Party of the Socialist Czechoslovakia).

9 The suspected assassination plan was reported by the investigative journalist Ondřej Kundra (magazine Respekt). It was investigated by the Czech Security Information Service. Russia denied the allegations. The report of the attempt and the ensuing conflict between Czechia and Russia resulted in the Czech government’s expulsion of two Russian diplomats. The assassination did not materialize in the end.

Appendix: Lag, cross-correlation, and associated KWs in numbers

Table

	Results from Companions		Association Arrays (results from MBA)
KW	Lag MS→ANTS	Cross-correlation	AA.ANTS.pre	AA.ANTS.post	AA.MS.pre	AA.MS.post
Lukašenko ‘Lukashenka’	1	0.92102	0	93	0	17
Bělorusko ‘Belarus’	2	0.88837	0	110	0	17
Běloruský ‘Belarusian (adj)’	0	0.86572	0	94	0	16
výbuch ‘explosion’	0	0.83048	0	2	0	5
Alexandr	0	0.82381	0	7	0	0
amonný ‘ammonium (adj)’	4	0.81107	0	0	0	0
dusičnan ‘nitrate’	4	0.78264	0	0	0	0
Minsk	1	0.76999	0	39	0	11
Floyd	3	0.76815	16	5	12	0
summit	1	0.74813	0	11	0	0
demonstrant ‘demonstrator’	0	0.74692	7	0	3	5
prezidentský ‘president (adj)’	3	0.74253	6	16	0	7
George	0	0.72617	0	1	8	0
koronavirus ‘coronavirus’	0	0.71684	228	96	165	114
rudý ‘red’	0	0.69709	42	0	0	0
protest	3	0.6938	20	52	14	14
Kalousek [Czech politician]	1	0.68769	0	2	0	0
volba ‘choice, election’	0	0.66630	10	50	7	30
nepokoj ‘unrest’	0	0.66506	2	2	0	0
požár ‘fire’	0	0.65528	0	0	0	2
epidemie ‘epidemic’	1	0.63382	84	7	25	0
rasismus ‘racism’	6	0.62122	1	25	0	0
pandemie ‘pandemic’	-1	0.60772	122	15	44	8
sněmovna ‘house [in the parliament]’	1	0.60440	2	1	19	0
Československo ‘Czechoslovakia’	0	0.60321	21	2	0	0
diplomat	0	0.59545	11	1	0	0
krize ‘crisis’	0	0.57758	55	9	2	0
volby ‘elections’	0	0.56683	0	0	0	0
protestující ‘protester’	0	0.55987	4	0	0	0
policejní ‘police (adj)'	0	0.53738	9	6	8	0
obnova ‘renewal’	1	0.52743	31	18	0	3
poslanecký ‘parliamentary’	1	0.52742	2	0	0	0
socha ‘statue’	0	0.52200	12	0	0	0
policista ‘policeman’	3	0.52078	20	14	11	1
zlato ‘gold’	-13	0.52058	0	0	0	0
válka ‘war’	0	0.51857	139	57	2	0
opozice ‘opposition’	0	0.51554	0	18	0	0
TOP [TOP 09, name of a political party]	2	0.50723	0	0	0	0
střet ‘conflict’	1	0.50251	6	0	14	3
sankce ‘sanction’	1	0.50101	0	37	0	0