1.2.1 Some Relevant Historical Background

Czechoslovakia, an independent country established in 1918 after the collapse of the Habsburg empire, was at first a country full of democratic hopes. The modern-day gender equality discussions, for example, have their main origin in the end of the nineteenth and the beginning of the twentieth centuries. The creation of a new country was a nationalistic project, and many women who wanted equal rights, or at least the right to vote, made those issues partly a nationalistic question (Feinberg Reference Feinberg, Fábián, Johnson and Lazda2022: 303) as a stance against the German-speaking empire. The idea that women are first and foremost mothers and wives was prevalent among both men and women (Frančíková Reference Frančíková2017: 8, 19, 27, 118) but was also seen as an argument to make women’s lives more equal when it came to the public sphere (Feinberg Reference Feinberg, Fábián, Johnson and Lazda2022: 305).

Explicit rights for women to be citizens of equal status with men, as well as all other residents irrespective of religion or origin, were eventually included in the first Czechoslovak constitution after World War I (Parlament Československé Republiky Reference Československé Republiky1920: Article 106). This was a clear success for the women activists, after the previous Austrian Civil Code where the husband was the head of the family (Feinberg Reference Feinberg2006b: 42–43), but the factual legal practice did not follow Article 106 to the word. The actual rights women eventually gained were mainly decided by the men of the commissions forwarding the law proposals to the parliament, which led to women still being bound by law to obey their husbands’ word regarding the household (51). In the Czechoslovak Republic, the more radical activists eventually had to compromise with the more traditionalist women of their organisations, and arguments about family and nation prevailed during the interwar period (70). Despite Article 106 of the Constitution and the fact that the women turned this into a nationalistic cause more than a gender issue, they eventually did not succeed in much more detail than other nations in Europe during this time.

Furthermore, this new state, later called the First Republic of Czechoslovakia, only existed for twenty years. In 1938, the country was further disrupted following Nazi Germany’s annexation of the Sudetenland in the north-west and Slovakia’s subsequent declaration of independence. Following World War II, Czechoslovakia fell under Soviet influence, leading to a communist coup in 1948 – despite the democratic elections before this coup giving the communist party around a third of the seats in parliament. This regime, which on paper promoted equality for all, eventually imposed several totalitarian principles and central planning, and suppressed dissent. The newspaper language – as all public language – was consciously used to pressure the readers in the ideologically sanctioned direction, to the extent that the newspaper sections, such as sport, foreign affairs and finance, were constructed in more or less the same way (see e.g. the chapters by Hedin, Woldt and Gammelgaard in Kress Reference Kress2012).

Another relevant note regarding both gender issues and occupational roles is that some women that supported the government of the Second Republic during the Nazi protectorate (1938–1945) defined the role women should have in society through the concept of ‘equality in difference’ (Feinberg Reference Feinberg, Wingfield and Bucur2006a: 100). This meant mainly accentuating women’s role in the home, but at the same time reiterating the idea that housewives performed manual labour, namely something that should be seen to be as difficult and creative as labour outside the home (Feinberg Reference Feinberg, Wingfield and Bucur2006a: 104–105). In many ways, this idea seems to be kept through both the first (democratic) and the second (protectorate) republics, and then reformed again, when in 1948 the communist party assumed power (Feinberg Reference Feinberg2006b: 206–207). Havelková notes (Reference Havelková2017: 40–41) that large efforts were made during the 1948–1989 period to relieve women of household work, but she also claims that these efforts never changed the role of men, and thus never removed the burden on women. She also claims (302) that by the end of and after that period, the further measures taken (such as increased maternity leave) only reinforced the role of women as wives and mothers and made it more difficult for women to pursue a career. This is reiterated in Weiner’s (Reference Weiner2007) sociological study on the 1948–1989 period, adding that women did not receive the same wages even for the same work (27) – something that may be expected from an ideology that claims that all output should be equally distributed. Issues of equality that were discussed in the West during 1948–1989 were by the communist party considered ‘bourgeois’ and thus ‘dangerous’ (Fojtová Reference Fojtová, Jusová and Šiklová2016: 112) – equality should be discussed in the way the communist party decided. For example, women were said to ‘help’ with – not participate in – ‘revolutionary tasks’ (Oates-Indruchová Reference Oates-Indruchová2016: 928).

The party’s expropriation (Havelkova & Oates-Indruchová Reference Havelková and Oates-Indruchová2015: 10–11) of the equality discourse became an obstacle to future work – after 1989, very few wanted to use terms or theories that were deemed contaminated by the former regime (see for example Nyklová Reference Nyklová and Kahlert2018). Cultural clashes often occurred when gender or feminist researchers from the West met with their colleagues in Czechoslovakia (1990–1993) or Czechia (after 1993). Some gender scholars, whether from the US or Europe, assumed their Czech colleagues would easily accept previous Western research, a phenomenon referred to as ‘westsplaining’. These Americans and Europeans had to accept the fact that the concept of feminism carried with it certain negative connotations in this part of the world (Fojtová Reference Fojtová, Jusová and Šiklová2016: 113–114; Jusová Reference Jusová, Jusová and Šiklová2016: 29–30).

However, opposition movements emerged, and in 1989, the communist regime eventually collapsed. Democratic presidential elections were held at the end of 1989, followed by parliamentary elections in 1990. The first president, Václav Havel, was a playwright and former dissident who wrote essays and plays about the totalitarian language used throughout this Communist society; how expressions that may for the uninitiated seem harmless are turned into despised expressions that make people want to turn their back on anything ‘socialist’. This is a reason for this study to start in this point in time. The president of the Czech Republic maintains some legal power (see e.g. Kubát et al. Reference Kubát, Mačutek and Čech2021: 140), and Havel was succeeded by economists Klaus (2003–2013) and Zeman (2013–2023). The latter became famous outside the country when he held up a toy gun with the inscription ‘For journalists’ during a press conference. The prime minister during most of Zeman’s years, Andrej Babiš, was often cited for promising to run the country like a business. The combination of Zeman and Babiš in the last decade of the analyses makes the time frame highly relevant for studies of representation in the news.Footnote ¹

1.2.2 The Czech Language

Czech is an inflectional, Slavonic language, with three genders (neuter, feminine and masculine), two numbers (singular and plural) and seven cases for nominal parts of speech (nominative, genitive, dative, accusative, vocative, locative and instrumental). This structure makes the word order more flexible than in languages such as English, since the cases explain which words go together within a clause or sentence. This also makes some research questions work better for studying the Czech language than they might for other languages. The features most relevant to this Element are the Czech language’s ability to place an adjective far from its noun without losing the connection, to create feminine forms of nominal words and to easily turn a word into its opposite by adding the negating prefix ‘ne-’. Following is an example, where the first adjective and its modified noun are underlined and the negation bold:

With the tools used, I intend to present methods that can be used for any language with the right resources. Furthermore, Czech is closely related to Polish, Slovak, Ukrainian and the other Slavonic languages, making this Element particularly useful for studies of and in those languages and countries.

1.2.3 Othering in the Czech Language

Othering in written Czech language has been studied by scholars such as Fidler (Reference Fidler2016) who divided the results into three historic parts: just after the fall of communism in the end of 1989, before Czechia joined the European Union (EU) in 2004 and after the country’s EU membership. She looked at collocations for the two words cizinec ‘foreigner, stranger’ and mluvící ‘speaker(s)’, and found, for example, a clear difference in the collocations of cizinec between the three periods. There was in her study, however, no enlarged discussion on the sentiments towards these strangers, providing another reason to analyse different ‘others’ in similar data.

The flexibility of the language also makes it possible to study othering by morphology, as in Thál and Elmerot (Reference Thál, Elmerot and Knoblock2022) who analysed the gendered endings given to transgender individuals in Czech online media and concluded that the clearest misgenderings were made deliberately. Using keymorph analysis (Fidler & Cvrček Reference Fidler and Cvrček2017) to study othering could be another option, but as yet there have been few attempts made for Czech (but see Fidler & Cvrček Reference Fidler, Cvrček, Fidler and 86Cvrček2018 on the difference in agency between Russia and Ukraine). An important fact for this Element is that, unlike some other languages, for example Polish, the use of the feminine form of a normally masculine noun is unremarkable (Hodel et al. Reference Hodel, Formanowicz, Sczesny, Valdrová and Von Stockhausen2017: 395). When an occupation has a male and a female form, such as malíř, malířka ‘male, female painter’, the female suffix in Czech mainly tells the reader something of the painter’s supposed gender self-identification.

1.3.1 The Subjectivity Lexicon for Czech

Not all languages are privileged to have any kind of subjectivity or sentiment lexicon available. For Czech, such a lexicon was published by Kateřina Veselovská (now Lesch) and Ondřej Bojar (Reference Veselovská and Bojar2013). This lexicon remains the most comprehensive source of evaluative language available for Czech. Its components, aims and usage are described in detail by Veselovská (Reference Veselovská and Bojar2013; Reference Veselovská2017). Its origins lie in an American equivalent (Wilson et al. Reference Wilson, Wiebe and Hoffmann2005) that has often been used in similar ways.

Lexicon-based sentiment analysis techniques are getting more and more popular (Benamara et al. Reference Benamara, Taboada and Mathieu2017: 202; Bestvater & Monroe Reference Bestvater and Monroe2023) within NLP, both inside and outside linguistics proper. One example from outside linguistics is Asgari (Reference Asgari2019), who used sentiment lexica from different languages to create a uniform model for sentiment analysis in different languages. He included the lexicon used in this Element (see Section 1.4) as one of the ‘gold standard’ tools to assess Bible translations and social media data (Asgari Reference Asgari2019: 171). A study within computational linguistics the same year comes from Zhao and Schütze (Reference Zhao and Schütze2019), who propose a ‘universal approach for sentiment lexicon induction’ (3514) by using a collection of similar lexica, including the Czech Subjectivity Lexicon. They do, however, conclude that several lexica needed manual adjustments or did not perform very well. Most of these and other sentiment lexica, as well as other tools, are based on modern-day reviews of entertainment websites and social media (Birjali et al. Reference Birjali, Kasri and Beni-Hssane2021). In contrast to the methodology employed in this Element, many of these studies on evaluation and sentiment use black box operations and are not very transparent.

The adjectives used in this Element are taken from the Subjectivity Lexicon for Czech and will be discussed further in the specific sections on occupations (Section 2.5) and gender (Section 3.3). Briefly, the adjectives, verbs and nouns in the Lexicon are classified into one of three ‘subjective’ categories: ‘positive’, ‘negative’ or ‘both’, depending on the subjective sentiment associated with them in Czech according to the researchers involved in that project. This is a research output that is useful if handled appropriately, and it is indeed a good start. Naturally, words alone do not cover the sentiment or tone of a whole text, which means that the results drawn from quantitative, semi-automated calculations must be compared to what was actually written. This is easily done with the material for this Element, since the corpus is a sub-corpus of the Czech National Corpus, freely accessible to all.Footnote ³

1.3.2 Problematic Issues with the Subjectivity Lexicon

In comparison to studies using all available adjectives surrounding certain nouns (in e.g. Islentyeva Reference Islentyeva2018, Zasina Reference Zasina, Fidler and Cvrček2018 or Černá & Čech Reference Černá and Čech2019), a problem for the Subjectivity Lexicon – as well as the semantic categories used in the gender section of this Element – is that its starting point was a translation of an equivalent lexicon for English. This translation included certain adjectives that should perhaps have been replaced by adjectives more frequently used in Czech text. A few examples are maximalizovaný from the English ‘maximise’, korozní from ‘corrosion’ and zmocněný, a translation of the verb ‘empower’, which only appear one single time each in the whole 1990–2018 Czech journalistic corpus. Another issue that had to be addressed early was the handling of negated adjectives. In Czech, both nominal and verbal word forms have an added prefix ne- in their negated form. The solution is that all negated adjectives in the dataset are classified as the opposite of their category in the Subjectivity Lexicon (e.g. the adjective nesolidární ‘lacking solidarity’ was classified as negative, because it is the negation of the positively classified solidární ‘having solidarity’). A third issue is the dominance of multifunctional adjectives. The most frequent adjectives overall from the Subjectivity Lexicon are velký ‘big, large’, dobrý ‘good’, rád ‘glad’Footnote ⁴ and poslední ‘last’. The latter adjective has caused problems for previous studies on the language used in Czech newspapers (e.g. Křen Reference Křen2017: 246–247). These four adjectives have sometimes been removed from the analyses that follow when they caused too much noise. Where that is the case, it is mentioned in the respective section. As Veselovská (Reference Veselovská and Bojar2013; Reference Veselovská2017) herself describes, there are more problematic issues, but most can be solved through qualitative reading of the actual source. More details are described in the following sections, but at the earliest stages of research design, a well-informed decision should be made about how to treat these kinds of issues.

1.3.3 Nouns with Adjectives

In this Element, evaluative adjectives that occur with specific sets of nouns are analysed as co-occurrences. The 753 adjectives of the Subjectivity Lexicon for Czech are classified into ‘positive’, ‘negative’ or ‘both’. Their co-occurrences in news media with nouns indicating nationality and religion have been analysed in two previous studies (Elmerot Reference Elmerot2021; Reference Elmerot, Filardo-Llamas, Morales López and Floyd2022), as summarised here. In these previous studies, the nouns were taken from public statistical lists of nationalities (Czech Statistical Office 2019) and state religion (United Nations Statistics Division 2020; Hackett & Grim Reference Hackett and Grim2012: 45–50;), as well as geographical and income-related classifications of countries (World Bank Group 2020). For Section 3, the female and male noun forms of these nationalities (e.g. ‘Malaysian woman’ or ‘Portuguese woman’) were extracted, as well as words for women, and their masculine equivalents, identified from the entry headed člověk ‘human’ from the most recent thesaurus for Czech (Klégr Reference Klégr2007). For Section 2, the main source of the nouns was the International Labour Organization’s ISCO lists of occupational titles (see International Labour Organization 2022), with additional occupational nouns from both the previously mentioned thesaurus and a list of perceived prestige of different occupations in Czechia (Tuček Reference Tuček2019). The latter two formed a ‘Non-ISCO’ group.

For simplicity, two distinct terms will be used in this Element: co-occurrences and modifiers. Co-occurrences are the combinations of nouns and adjectives that go together in the same text, even if the adjectives might describe something in the context other than the people. In this Element, modifiers refer to the adjectives that appear just beside the nouns and, in these analyses, are prominent enough to affect the overall evaluation in a certain year or period. Modifiers may therefore include both attributive and predicative adjectives. The idea is thus to include both modifiers and co-occurrences, as both contribute to the discourse. For example, if ‘nurse’ is found one or seven words to either side of ‘paranoid’, both co-occurrences are classified as negative. However, the co-occurrences that are further apart, and thus not modifiers, are given less weight in the analyses. More details are found in the respective sections.

1.3.4 The Corpus

The co-occurrences of the previously mentioned nouns and adjectives were extracted into a dataset from the journalistic sub-corpus of the SYN (as in ‘synchronic’) series (release 8) of the Czech National Corpus, containing 4,499,370,372 (≈ 4.5 billion) tokens (running words, excluding punctuation) from approximately 200 titles. The SYN series are updated on a yearly basis, which facilitates repeating a study with newer material. This sub-corpus contains newspapers and magazines from 1990 to 2018, thus covering the period from the first democratic elections after the communist, Soviet-led regime to a new time of economic ups and downs, populism and migration-related debates. A multi-dimensional text study of genres based on the same corpus (Cvrček Reference Cvrček2022) argues that these news outlets have gone through a shift from general to specific descriptions of reality, reducing the degree of cohesion and retreating from ‘stated attitudes’. His study at the same time rejects theories about a tabloidisation of the news press during the same period as that discussed in this Element (32). The majority of the titles in the corpus are mainstream press, such as the dailies Lidové noviny and Mladá Fronta DNES and the magazine Respekt, but there are also a few titles that may be considered alternative press. One such is the Parlamentní listy, that began in print in 2003 and may have been a way for politicians to interact directly with their voters. Today it is better described as a nationalistic and socially conservative outlet of many re-used stories, with little journalism involved (Gregor & Mlejnková Reference Gregor and Mlejnková2021; Štětka & Mihelj Reference Štětka and Mihelj2024: 188–189). Since it is a very popular outlet, it still merits its place in a corpus of what Czechs normally read. This journalistic corpus is thus an ‘acceptably representative sample of a population of language users, a language variety, or a type of discourse’ (Ädel Reference Ädel, Paquot and Gries2020: 4).

1.3.5 The Constructed Dataset

The entire dataset constructed for this and the two previously mentioned studies has about 33 million observations or sampling points from the years 1990–2018, where each observation includes one adjective co-occurring with or modifying one noun in one issue of a certain newspaper or magazine. Table 1 illustrates the observations using two examples: the noun Afričan with the adjective dobry, and the noun Bulharka with the adjective zkušený. The table shows the sentiment of the adjective, whether or not it is negated, and the frequency. When there are multiple co-occurrences of the same noun and adjective (as in the afričan example in Table 1, shown in the ‘Fq’ column), the variable ‘adjusted frequency’ (AdjFq column in Table 1) measures the average word distance between the nouns and adjectives for that text.Footnote ⁵ The idea is to measure the adjective’s possible influence on the noun. In Table 1, Afričan ‘African man’ and dobrý ‘good’ co-occur several times in the same issue, but the fact that the adjusted frequency is lower than the raw frequency tells us that the adjective is not often a modifier to the noun. The second row, in contrast, shows Bulharka ‘Bulgarian woman’ and zkušený ‘experienced’ occurring once next to each other, giving an AdjFq of 1.

The adjusted frequency (AdjFq) is used instead of doing a search with pre- or post-modifying adjectives, based on proximity findings by Cvrček (Reference Cvrček2014). There may well be more than one word between the adjective and its modified noun and, in addition, adjectives in the near vicinity may be highly relevant for the analysis; see Table 2.

Example 1 in Table 2 does not contain a modifying adjective for uklízečka, but the noun is close to a modifier (placené ‘paid’) of another noun (společnice ‘female companion’) of a similar category, in this case another low-paid occupation. If this had been a modifier included in the Subjectivity Lexicon, its classification there would have made this a classified co-occurrence. Example 2 does include an adjective included in the Subjectivity Lexicon, správná ‘proper’, modifier of the noun uklízečka. This is then classified as positive, according to the Lexicon’s classification. Example 3 has three adjectives that all modify the noun: mladičkou ‘very young’, sympatickou ‘nice’ and rozesmátou ‘smiling’. All three are included in the dataset, with slightly different adjusted frequency values due to their different distances from their noun. Example 4 has two adjectives (ošklivá ‘ugly’ and zakomplexovaná ‘full of complexes’) modifying the unknown subject of the sentence. Here, the noun uklízečka is a predicate nominative referring to the same subject, explaining what she did for a living (pracovala jako ‘worked as’). They are also included and classified, but with different adjusted frequency values.

Before the quantitative analyses were performed, two more variables were added. First, a variable called sentiment value was added, where a positive classification from the Subjectivity Lexicon was turned into +1, a negative classification to −1, and the occurrences where the adjectives were classified as both became 0. This value was then multiplied with the adjusted frequency for each observation, and the result is a variable called adjusted sentiment value:

$\text{adjusted sentiment value }=\text{ sentiment value }\times \text{ adjusted frequency.}$

The lower the adjusted sentiment value, the lower the impact on the results. This adjusted sentiment value is then summed and shown per co-occurrence (of adjective and noun) per year to reveal the difference between the positive and negative classifications. In two previously published studies (Elmerot Reference Elmerot2021; Reference Elmerot, Filardo-Llamas, Morales López and Floyd2022) the calculations were based on the mean of this adjusted sentiment value per year, meaning the results differ depending on how many published issues each source (newspaper or magazine) had in the corpus each year. With the summed value divided by the actual number of tokens in the corpus each year, as in this Element, the positivity and negativity are recorded accurately no matter how many issues were included. The results of these calculations give an overall insight and can then be taken as a basis for a study that would focus on modifying adjectives.

1.4.1 Summary of the Othering and Stratification of Nationalities

During the Cold War, from 1948 to 1989, the Czechoslovak Republic was a part of the Soviet satellite countries. All countries behind the Iron Curtain were seen as sibling countries and neighbours. However, it now considers itself a part of the ‘Western’ world, together with the rest of the countries in the European Union (EU) and the North Atlantic Treaty Organisation (NATO). This change is reflected in the representation of other countries in the news media. The media uses similar amounts of positively classified adjectives to describe Czechs as it does for other high-income countries, according to the World Bank’s classification. Generally, the adjectives in the Subjectivity Lexicon for Czech provide a clear stratification of the four income groups in that classification. The news press tends to use more negative adjectives to describe lower national income groups. When combined with these graphs, the qualitative analysis from previous studies shows that Arabs, Muslims, and particularly Afghans are the most distinct out group in the current millennium.

2.4.1 Overall Evaluation

Figures 5–6 in this section are comparisons of only the positive and negative classifications to see which is dominant within the evaluated co-occurrences. Adjectives classified as both positive and negative, thus rendering a 0 as the variable value, are excluded from these figures. The sum of the absolute numbers for the two classifications per year are displayed using contrasting colours, bright green for positively, and dark red for negatively evaluated co-occurrences. If a group had an equal amount of positively and negatively evaluated co-occurrences, the green and red would meet at the black horizontal line.

To determine the most negatively represented occupational groups in total, Figure 5 displays the proportion of positive and negative modifying adjectives used for each group during the entire period. The x-axis indicates the ISCO group, and the y-axis indicates the proportion of evaluation. Figure 5 is sorted by the total amount of negative classification.

Figure 5 Proportion of negativity and positivity per occupational group for the whole period.

There are indeed clear differences between the groups in Figure 5, and there is neither a direct nor inverse relationship between the Subjectivity Lexicon’s evaluation and the hierarchy of ISCO groups 1 to 9. Managers (ISCO 1), Professionals (ISCO 2)Footnote ¹⁵ and Technicians and associated professionals (ISCO 3) do have a similar overall amount of positivity, but Craft and related trades (ISCO 7) have an overall evaluation that is more similar to ISCO groups 1–3 than to groups 8 or 9. In comparison to ISCO groups 7 and 9 (Craft and related trades and Elementary occupations), Service and sales (ISCO 5) receives more negative evaluations. Additionally, in one year, the Service and sales group received more negativity than positivity, although the corpus for that year was rather small.

To show possible changes over time, the proportions per individual year are calculated. The results are presented in Figure 6.

Figure 6 Proportion of negativity and positivity per occupational group and year.

Only one group, Plant & machine operators with assemblers (ISCO 8), received more negative than positive classification for more than one year. All other groups received more positive than negative classification over time, except for Service and sales (ISCO 5), where negativity outweighed positivity in 1994. The non-socioeconomic groups Military (ISCO 0) and Public officials (non-ISCO) appear similar to the Clerical Support (ISCO 4) and Service and sales (ISCO 5) groups respectively. ISCO 0 and ISCO 4 exhibit higher positivity at the beginning of the timeline compared to the end, while the other groups show the opposite trend. This order may be partly explained by the theory of supervision as a factor (van Leeuwen et al. Reference van Leeuwen, Maas and Miles2002a: 52–53). The group with the most negatively classified co-occurrences for the entire period consists of occupations taken from the prestige list that belong to public officials or company owners – people to whom supervisors report, rather than the supervisors themselves. Following these are groups that are subject to some level of supervision, including military staff, clerical support workers and machine operators. The co-occurrences with the most positive classification throughout the entire period are found in groups where individuals may supervise others, such as managers and other types of specialists.

Figures 5 and 6 demonstrate that there is no precise correlation between socio-economic status and the extent of positive evaluation. An alternative grouping that better follows the data for this section is shown in figures 7–9. There, similar professions are grouped based on their frequency of mention and proportion of positive and negative ratings. For further analysis, three groups have been identified:

Figures 7–9 visually represent the evaluation of co-occurrences and their frequency relative to the total number of tokens in the original corpus. The y-axes of the charts represent a value obtained from twice the summed adjusted sentiment value, divided by the number of tokens per year. This is because each newspaper or magazine is published with a different number of issues each year. Figures 7–9 show the number of mentions for each occupational group within the groupings, represented by the length of the bars. The figures also display the normalised distribution of positive and negative evaluations, represented by the proportion of each bar above or below the central horizontal line. The scales differ between the positive and negative facets of the figures due to variations in the strength and frequency of positivity and negativity for different occupations. Figure 9 shows that the Clerical support group (4) has the longest bars in most years, indicating that there are more evaluated nouns in the news from this group than from the other groups in the same graph. Additionally, Elementary occupations (9) often have the fewest mentions in both facets, but on the right-hand side, in the last two years, they have more positive co-occurrences in the news than any other group in this facet.

Figure 7 The occupational groups with higher presence and their relative evaluation. Y-axis numbers are relativised by the total number of corpus tokens.

As previous research shows, the Professionals or Specialists (ISCO 2) – including occupations such as journalists, physicians, actors, judges, and singers – is mentioned more frequently than other groups. This is true for all years in positive co-occurrences and for most years in negative co-occurrences. The coefficient of variation, which is a relative measure of dispersion, is a useful tool for validating the results presented here. It is calculated by dividing the standard deviation by the mean of the relative sentiment value for each group. The results show that the Professionals (2) are the most stable group over time. Notably, the non-ISCO group, mainly composed of politicians, has the highest normalised number of negative evaluations, as demonstrated by the longest negativity bar in Figure 7. Despite this negativity score (−40), there is also a high positivity score (+52), which means they get an overall positive evaluation.

Figure 8 shows that the Service and sales group (ISCO 5) has positive y-axis numbers similar to those of the Managers group (1), while the negative y-axis numbers are close to those of Technicians, etc. (3) from figure 7. The Military group (0) and the Craft and related trades (ISCO 7) are most often the least occurring in this graph. At face value, the Military group appears more frequently at the beginning of the timeline, even when their numbers are relativised by the actual number of tokens in the corpus. However, soldiers are often featured in news stories related to other countries. Closer examination of the corpus reveals that prior to 2004, there were wars in Yugoslavia and debates about the Czech(oslovak) Republic(s) sending troops there. There were also debates focused on military issues before Czechia joined NATO in 1999. After 2004, there are more stories about older battles and military events. References to the extremely popular novel The Good Soldier Švejk are another reason for the high frequency of the occupation name ‘soldier’.

Figure 8 The occupational groups with medium presence and their relative evaluation. Y-axis numbers are relativised by the total number of corpus tokens.

Figure 9 illustrates that towards the end of the analysed timeline, these four groups exhibit similar values. However, Clerical Support (ISCO 4) has significantly more positive co-occurrences at the beginning of the timeline, and the Elementary Occupations (9) receive more positive co-occurrences towards the end. The Skilled agricultural, forestry and fishery workers (6) are evaluated the most in the middle part of the time period for both facets. The Plant and machine operators (8) and Elementary occupations (9) show similar levels of positive evaluation, with their bars often approximately equal in height. However, the Elementary occupations have a longer positivity bar and there are slightly longer negative bars for the Plant and machine operators.

Figure 9 The occupational groups with the lowest presence and their relative evaluation. Y-axis numbers are relativised by the total number of corpus tokens.

2.4.2 Results of the Overall Evaluation

When studying frequency of evaluation in the news, it seems the previous research which collapses the ISCO groups into different overarching groups does not hold. Katrňák (Reference Katrňák2012) used a minimum of three (manual workers, non-manual workers and farm workers) or five groups (non-skilled, skilled and farm workers, as well as white collar and ‘petite bourgeoisie’) (683, compiled from Erikson and Goldthorpe Reference Erikson and Goldthorpe1992: 39), whereas we observe a different grouping here. The Clerical support workers (ISCO 4) cannot, for example, be grouped with manual workers, whereas some occupational titles in the Service and sales workers could be included in a broader group of non-manual workers. The theory on supervision taken from the HISCO classifications (van Leeuwen et al. Reference van Leeuwen2002b) may provide a better rationale for the frequency of evaluation in the media: the groups that receive the most media attention and evaluation are Public officials etc., Professionals, and Technicians on a higher level, who are typically in supervisory roles.

Another reason could be service to and contact with other people. The middle group includes managers, who also supervise but may not always be appreciated, as well as the police and military, who supervise and give orders to the general public. All these professions involve close contact with the public, and the police at least have a role of serving the general public (Zákon č. 273/2008 Sb.).Footnote ¹⁶ While the supervision theory does not provide an explanation for why Craft and related trades (7) fall into this middle group, the service theory may offer some insight. It is important to note, however, that the frequent appearance of both surnames and famous deceased painters in this group may be the reason for its frequent media coverage. The media co-occurrence groups at the lowest level do not include occupations where supervision is an obvious part of their occupation, and they may often work far from other people. A combination of the theory of supervision with a theory about occupations serving others may thus be an important factor for a group to be more frequently mentioned with evaluating adjectives in the media.

2.5.1 The Higher Groups: Public Officials etc., ISCO 2 and ISCO 3

The Public officials, etc. group consists of ten titles, such as ‘mayor’, ‘minister’, ‘priest’ and ‘ambassador’, plus the non–public official nouns majitel ‘proprietor, owner’ and stavebník ‘builder, constructor, developer’.Footnote ¹⁷ The Professionals group, ISCO 2, comprises the widest range of occupations, from physicians and nurses of various kinds to journalists, actors and teachers. Since journalists are included, the group’s sheer frequency may be seen as a verification of the ‘like me-hypothesis’, that journalists mainly write about people in the same socio-economic group. The ISCO 3 group, Technicians and Associate Professionals, is in this dataset – despite its name – dominated by sports-related occupations from 1993 onwards. Additionally, it is one of the most positively evaluated groups in Figure 5.

In this Higher frequency section, the evaluative adjectives often do not evaluate the occupations themselves. In certain cases, they may instead assist in creating a discourse prosody (Stubbs Reference Stubbs2002: 65) – an evaluative discourse unit. A clear example of this is the adjective předchozí ‘precedent’. In more recent years, this adjective has a rather high co-occurrence with majitel ‘owner’, whereas in the 1990s, that noun tended to co-occur with cenný ‘registered’ or with ochranný ‘protective, safety’, mainly due to the bigrams cenný papír ‘security, financial instrument’ and ochranná známka ‘trademark’. This indicates the significance of these new concepts in post-communist government discourse, as discussed in newspapers at the time. Another example of evaluative prosody is the noun lékař ‘physician’, with frequent co-occurrences in the top five negative evaluations, largely due to its association with the adjectives nemocný ‘sick’ and zraněný ‘hurt’. However, the doctors themselves are frequently described as both neschopný ‘incompetent’ and schopný ‘competent’, with the former being one of the few negated adjectives in the top list.

A previous study on the frequency of certain occupations in the Norwegian media found that construction workers and ‘caring’ occupations were described less frequently (Figenschou et al. Reference Figenschou, Eide and Einervoll Nilsen2021: 84). In this dataset, the only construction-related occupation that is clearly declining in frequency is stavebník from this group, which refers to builders, constructors and developers. The same phenomenon cannot, however, be seen for any kind of nurse in this dataset (there are four in total: sestra, zdravotní sestra, vrchní sestra and staniční sestra, in order of relative frequency). When their absolute frequency is normalised by the tokens in the corpus each year, their presence remains stable over time.

The evaluative development of journalists should be examined, since journalist is one of the nouns in the Prestige list that have experienced a decline in popularity in 2004–2019 (see Table 4). The dataset does not reflect the loss of prestige mentioned, except for 2018, where there is a sudden increase in negative co-occurrences for both male and female journalists, indicated by the negative adjective zabitý ‘murdered’, which this year co-occurred with journalists more frequently than any other year. This was due to the murders of two journalists in the neighbouring country of Slovakia and of a journalist at an embassy in Turkey. These incidents caused a stir in several European countries. Again, this shows that using pre-classified variables makes the need for close reading very important, as the journalist is described as having undergone a negative event, rather than behaving negatively.

Some examples of peculiarities indicating the relevance of close reading have been mentioned, but others may also be relevant for the design of future research. Among the top five most frequent positive co-occurrences in this Higher group, two are used with female nouns: sestra ‘nurse’ and učitelka ‘teacher’. However, the most frequent, positively classified adjective for the latter noun is základní ‘basic’, which refers to primary or elementary school. The categorisation of adjectives is another peculiarity: sociální ‘social’ is categorised as a positive adjective, which leads to some problematic positive classifications. One example is the phrase ministr(yně) práce a sociálních věcí ‘(female/male) minister of work and social affairs’. Sociální is the most influential positive adjective in 1995, and the strongest driving force of the overall positivity throughout the whole timespan. Another positive adjective that also dominates in more recent years is nezávislý ‘independent’. This is partly used about mayors who are not affiliated with a specific political party. However, since 2004 Czechia has had a political party called Starostové a nezávislí ‘Mayors and Independents’,Footnote ¹⁸ which has also created some bias in this dataset. The negative categorisations present a similar issue, as the adjective stínový ‘shadowy’ consistently appears alongside the most frequent nouns ministr/ministryně ‘male/female minister’ throughout the entire time span, rarely describing any other group. This can be explained by the existence of a ‘shadow cabinet’ of ministers. The masculine form is one of the most prominent negative co-occurrences in both 2001 and 2013, with female ministers ranking 5th in 2001 after the formation of a special shadow government of women (see Section 3) and 13th in 2013.Footnote ¹⁹

One of the groups, ISCO 3, is here dominated by sports, which is a large part of news media. The co-occurrences here most often represent the emotions of the occupation, rather than the evaluation. The noun kouč ‘trainer, coach’ appears in the top five with both negatively and positively classified co-occurrences. This noun appears more often in the negative co-occurrences, starting in 1998, and for four years only with the adjective poražený ‘defeated’. In more recent years, the emotions behind these defeats have become more apparent, as the adjective zklamaný ‘disappointed’ is highly frequent among the negative co-occurrences. Among the positive co-occurrences, the noun appears in the top every year since 1999; in the top five only with the adjective zkušený ‘experienced, skilled’.

The analyses also point to relevant results for writing about professions in a balanced or equal manner. When examining which adjectives most commonly modify the teacher nouns (not just co-occur with them), there are some differences between genders. The teacher nouns in masculine form are more frequently modified by the adjective kvalifikovaný ‘qualified’, which is the fourth most common adjective for them after the multifunctional základní ‘basic’, dobrý ‘good’ and velký ‘big/great’. In feminine form, teacher nouns are often modified by zkušený ‘experienced, skilled’ (fifth most common modifier for both genders), but even more commonly, they are described as hodný ‘nice, kind’ (see Section 3.3.3). In contrast to previous research on teachers, the two adjectives in the Subjectivity Lexicon that may be translated as ‘caring’, laskavý and starostlivý, do not rank among the top fifty most frequent modifiers. Laskavý is present in the top 100 most frequent modifiers (at position 25 for female teachers and 66 for male teachers), but starostlivý is less frequent. Another adjective found in previous research on teachers is ‘honest’. In this dataset, both čestný and poctivý ‘honest’ are used less frequently than ‘caring’ in reference to female teachers, but more frequently in reference to male teachers. Čestný is never used in its negated form, and nepoctivý ‘dishonest’ only once in the whole period, in the near vicinity of nouns referring to male teachers.

It is challenging to summarise these three groups, given the extensive scope of their representation within the dataset. A significant proportion of these occupations are perceived as supervisory, and the majority of them involve providing services to others. This may be a contributing factor to the high level of positivity associated with these roles, despite the prevalence of negative associations, such as those related to illness and mortality, in occupations within the care sector. A closer examination of the individual co-occurrences reveals the limitations of using a sentiment lexicon. In some instances, the ‘positive’ adjectives are descriptive rather than evaluative, as in the denotations ‘qualified teacher’ or ‘elementary school teacher’. In other cases, such as ‘journalist’ and to a certain extent ‘physicians’, the negative adjectives used do not reflect an evaluation of members of the profession. This illustrates that a quantitative study alone cannot provide an accurate representation of an evaluative attitude, but it can serve as a valuable foundation for sentiment analysis.

2.5.2 The Middle Groups: ISCO 0, ISCO 1, ISCO 5 and ISCO 7

ISCO 0 is slightly different from the others, in that all professions are military. For this dataset, only the ones not present in other categories were included, and the two most frequent nouns are voják ‘soldier’ and generál ‘general’. ISCO 1 comprises in this dataset fifteen different managerial nouns, but only three of them, ředitel ‘director, head (of something)’, manažer(ka)Footnote ²⁰ ‘manager’ and inženýr ‘engineer’ (also an academic title, Master of Science) are mentioned frequently enough to make it into the top ten of the most evaluated and frequent co-occurrences per year. ISCO 5 has one dominant noun, policista ‘police officer’ with 34 per cent of the co-occurrences, followed by hasič ‘firefighter’ with 19 per cent and strážník with 11 per cent. Some years, the noun strážce ‘guard’ also is quite frequent. The noun with a third position in the dataset, strážník ‘constable, warden’, is sometimes close to being a synonym for a police officer.Footnote ²¹ These are represented as dangerous occupations, which is why this ISCO group is distinctive: it is one of the few to have more negative than positive evaluations in at least one year, and the negative evaluations have been as high as the positive for several years, as shown in Figure 5. Finally, ISCO 7, a group of Craft and Related Trades occupations, is dominated by the noun malíř, which can be translated as ‘painter’ or ‘artist’, but may also refer to a ‘decorator’ or ‘house painter’. Therefore, it is a broad term that encompasses both blue-collar workers and renowned artists such as Toyen, the Czech painter whose paintings sell for millions of both Czech crowns and British pounds. This scope explains why it has more than twice the number of occurrences compared to the second and third most frequent nouns in this group, kovář ‘(black)smith’ (four out of six occurrences are surnames) and krejčí ‘tailor’ (almost nine out of ten are surnames). The fact that many of the nouns in this group are surnames again highlights the importance of closer scrutiny when using methods that cannot distinguish them. The analyses that follow focus solely on the actual occupations in the given context.

The group of craftsmen and related occupations is the only one of the Middle groups where the adjectives most often evaluate the occupations and not the discourse, or act as explanatory modifiers. The most influential co-occurrences in the positive classification are slavný ‘famous’ and významný ‘important, significant’ with malíř ‘painter’, which is one reason the ISCO 7 group’s positivity is higher compared to the ISCO 6 and 8 around it. This is also enhanced by the normally rarer adjective vynikající ‘excellent’, which is the ninth most frequently used adjective in this group that modifies the noun, including the multifunctional velký ‘great’, dobrý ‘good’, and so on. In these most driving co-occurrences, we also find kvalifikovaný ‘qualified’ and tradiční ‘traditional’ with řemeslník ‘craftsman, artisan’, a noun whose eight out of ten modifiers are positively classified adjectives. Among the negative rankings for ISCO 7, we find several co-occurrences with the noun silničář ‘road worker’. Apart from the multifunctional adjectives, the most frequent adjectives with road workers are špatný ‘bad’, těžký ‘heavy, hard’ and silný ‘strong’. To a lesser extent, there is also the adjective poškozený ‘damaged, impaired’, which mainly concerns the roads they work on. Another noun recurring in the top five list of most common, negative co-occurrences is kominík ‘chimney sweep’, modified by the adjective falešný ‘fake, false’. There have been several occasions where fraudulent chimney sweeps roamed the Czech countryside to earn money, which is reflected in the frequencies of this dataset and gives a negative representation of the occupation. One newspaper concluded in a subtitle that Znát svého kominíka se vyplatí ‘It pays to know your chimney sweep’,Footnote ²² meaning that otherwise, a fraudulent one may steal your money.

The Military group is decreasing in prominence after the 1990s, and in the three years 1991, 2008, and 2018, when the negativity and positivity were almost equal, the most common noun voják ‘soldier’ most frequently co-occurs with zraněný ‘injured’ or zabitý ‘killed, which are indeed often modifiers of the noun. However, the negatively classified adjective válečný ‘war-’ also adds to the general negativity for this group. It is noteworthy that the adjective sebevražedný ‘suicidal’ is co-occurring more with the noun voják ‘soldier’ (965 times) compared to any other occupational noun (top two, policista ‘police officer’, has 669 and top three, ministr ‘minister’ has 256 co-occurrences, of which only a handful or two are actual modifiers). The highest frequency of sebevražedný + voják as a proportion of all tokens in the corpus was in 2018, followed by 2014 and 2003. Upon closer examination of the corpus, it becomes apparent that this is often related to suicide bombings or attacks in which soldiers are involved, typically because the target is a military object. The negative discourse prosody (see Section 1) about soldiers is thus increased by these co-occurrences, even though they are not suicidal themselves. On the other hand, the highest positivity for this group is created by the adjective základní ‘basic’ or ‘fundamental’, which has been classified as positive in the Subjectivity Lexicon. It often co-occurs with voják because of the collocation vojáci základní služby ‘soldiers in [basic] military service’, which is a neutral representation rather than a positive one. Among the other positive adjectives co-occurring with the noun ‘soldier’, we also have mírový ‘peace(ful)’ and bezpečnostní ‘secure, safe(ty)’, mainly due to the presence of international UN peacekeeping forces and other peace missions. An example: Naši vojáci působili v rámci mírových sil OSN v Mogadišu od 23. července. ‘Our soldiers have been operating in the framework of the UN peacekeeping force in Mogadishu since 23 July.’ Both ‘peace(ful)’ and ‘secure, safe(ty)’ put the soldiers themselves in a more positive light on a wider dicourse level.

The most frequently occurring female nouns in this group are manažerka ‘manager’ and vrchní sestra ‘charge nurse/head nurse’ or ‘nurse practitioner’ for ISCO 1 and for ISCO 5 modelka ‘female model’ (classified as an ‘Other Sales worker’ in the ISCO descriptions). For the female manager, seven adjectives are equally frequent, seen throughout the whole time: příjemný ‘nice, pleasant’, schopný ‘able’, ‘competent’, sociální ‘social’, špičkový ‘peak, high-’, tvrdý ‘hard’, úspěšný ‘successful’ and významný ‘important’. Except for the first, they are all adjectives that occur more often with men than women, as seen in Section 3. It is noteworthy that female managers are frequently represented as schopný ‘able’, ‘competent’ more than any other occupation where either a feminine noun is present in the dataset or women are known to be prominent, such as with teachers. For nurses, the most frequent adjectives are sociální ‘social’, nemocný ‘sick’ and rehabilitační ‘rehabilitative’, but also záchranný ‘salvage/safety’, možný ‘feasible’, špatný ‘bad’ and příjemný ‘nice, pleasant’. The noun modelka ‘female model’ most frequently (and unsurprisingly) co-occurs with krásný ‘beautiful’, but also with other mainly positive adjectives such as slavný ‘famous’, úspěšný ‘successful’ and půvabný ‘charming’. These co-occurrences are often (41 per cent) found in magazines, which means that a study only analysing daily newspapers would yield different results for the models. It is worth noting that schopný ‘able’, ‘competent’ and možný ‘feasible’ (but also sometimes ‘plausible’ and ‘possible’) are adjectives that are more frequent with men in general but rank highly (number five with nurses) when referring to women’s occupations in these Middle groups. A comparison with the dataset used in Section 3 reveals that manažerka ‘female manager’ is indeed one of the absolutely most common female nouns modified by these adjectives.

Summarising these occupational groups, positive adjectives are mainly associated with artistic and handicraft-related occupations such as the nouns malíř ‘painter’ and řemeslník ‘craftsman, artisan’, and to a certain extent peacekeeping military operations, but also with competent female managers. The more negative adjectives are associated with more dangerous, heavier or dirtier occupations such as the nouns voják ‘soldier’, policista ‘police officer’ and silničář ‘road worker’. At the same time, for the more frequently represented occupations in the dataset, the adjectives less often describe the occupations, and more often the discourse around them, such as válečný ‘war-’, or are explanatory modifiers, such as záchranný ‘rescue-’ in ředitel záchranné služby ‘director of (the) rescue services’. In this case, the theory that service and supervisory functions would give a more frequent and positive media representation seems confirmed by the positive co-occurrences, while the negative adjectives more often concern the context more than the professions themselves.

2.5.3 The Lower Groups: ISCO 4, ISCO 6, ISCO 8 and ISCO 9

The following paragraphs concern the least prominent groups in the data. In these low-frequency groups, it can be noted that the adjectives more often modify the nouns directly rather than appearing as co-occurrences nearby. ISCO 4 stands out here, partly because clerks and similar support occupations are normally part of a higher social group than occupations from ISCO 6, 8 and 9, but also in a purely linguistical manner: the occupations of ISCO 4 are not also surnames.

The ISCO 4 group is dominated by the noun úředník ‘clerk’, which has more than twice as many co-occurrences as the second most frequent noun, prodejce ‘salesperson, seller’. This group receives increased negative classification as time goes by, and in the two most negative years of the 2010s, the salespeople nouns are the most prominent among this group. This is mainly due to a debate on door-to-door salespeople, pedlars. At the time, this practice was still rather common, but there were calls for an amendment to the law. This resulted in media examples such as Doplácíme na nekalou praxi nepoctivých prodejců dek či hrnců. ‘We are paying the price because of unfair practice conducted by dishonest sellers of blankets or pots.’Footnote ²³ The clerk nouns also receive a large share of the negative classification due to the modifier nízký; in its comparative form, nižší úředník means junior clerk. But almost as common is the adjective špatný ‘bad’, followed in frequency order by podezřelý ‘suspicious’ and obyčejný ‘ordinary’, which all are frequently used throughout the period. On the other hand, the positivity accompanying clerk nouns, which is larger during the first years, seems to be focused on fewer adjectives. Schopný ‘able, competent’ is the most common, closely followed by možný ‘feasible, possible’ and the multifunctional sociální ‘social’. These adjectives appear consistently throughout the period. Some of the seemingly positive schopný are, however, in a doubtful or even negative context. An example from the transition from communist times to capitalist times is an interview in Mlada Fronta Dnes with the Plzeň boss of the vehicle manufacturer Škoda who was quoted saying Například z více než 1300 úředníků zbude jen 60 schopných, opravdu řídících pracovníků. ‘For example, out of more than 1,300 clerks, there will be only 60 capable workers left that are truly executives.’Footnote ²⁴

In the other less frequently mentioned groups, a methodological problem arises with surnames, such as Chmelař ‘hop-grower’, Horník ‘miner’ or Kočí ‘coachman’. Upon closer examination of the corpus concordances, it becomes clear that, for example, at least half of the co-occurrences with the possible hop-growers actually refer to someone with that surname. For ISCO 8, the most commonly used modifier is connected to a surname: boží ‘godly’ or ‘divine’ is simply a geographical modifier of a politician, Jan Horník from the city Boží Dar, who has been active throughout the entire period covered by this Element. Out of the 71,000 mentions of this noun in the entire journalistic corpus, approximately 20,000 are about him or other Horníks and need to be ignored. Most of the group’s most prominent nouns are, however, not names, and after some controls for which lemma is actually used, the results presented here are deemed valid.

Figure 6 shows a negative evaluation of the ISCO 6 group of Skilled Agricultural, Forestry and Fishery Workers initially, but recent years have seen a 70 per cent positive classification. The top five co-occurrences per year indicate a clear transformation of the country from a communist regime with collective farms. During the 1990s, the noun zemědělec ‘farmer’ frequently appeared in the top five most frequent positive co-occurrences, but after 2003, it drops to lower frequencies. In the list of the top five most driving negative co-occurrences, zemědělec ‘farmer’ appears in four out of the top five positions during nine of the 29 years, as recently as 2012, and occupies at least two of the top five every year. Negative co-occurrences with zemědělec ‘farmer’ are less frequent than positive, and mostly relate to harvests. The adjective nízký ‘low’, followed by špatný ‘bad’, refers to events such as extreme weather that is harmful to farmers. Another common negative term is postižený ‘affected’, used in a similar context (e.g. … pomoc zemědělcům postiženým povodněmi … ‘ … aid to farmers affected by floods … ’) which was very frequent in 2002. The negatively classified poslední ‘last’ appears frequently in a climate context to compare one year with previous years. As an experiment, I removed the ambiguous adjectives nízký ‘low’ and poslední ‘last’. Consequently, the negative classification decreased significantly for the group (approximately 25 instead of 30 per cent over the last five years). I subsequently removed all three ambiguous adjectives (nízký ‘low’, poslední ‘last’ and vášnivý ‘enthusiastic’), which did not change the ISCO group’s evaluation significantly; the evaluation became only slightly less negative than in Figure 6.

In more recent years the most frequent and driving, positively classified, co-occurrences are instead dominated by the noun rybář ‘fisher’. In the top five, the fisher nouns are modified with úspěšný ‘successful’ and zkušený ‘experienced’, but most of all with vášnivý ‘passionate, enthusiastic’. This adjective is not commonly used to describe other occupational nouns. It is worth noting that these positively classified texts most often refer to people who fish in their spare time, rather than professional fishermen.

The ISCO 8 group has the title Plant and Machine Operators, and Assemblers, but also includes some other types of ‘operators’, generally considered drivers. The most frequent noun overall is horník ‘miner’ (which is a surname in several cases, as described earlier), followed by taxikář ‘taxi driver’. Among the positive adjectives, a religious adjective appears among the most frequent co-occurrences, namely svatý ‘holy’. Due to the high risk involved in mining, there are many saints for both miners and mining, and their images or statues are often present in mining workplaces.Footnote ²⁵ While these saints are reflected in the news and contribute to the context, they do not directly modify the miner nouns. Balancing this positivity are adjectives about the dangerous work in mines. In 1992, horník ‘mineworker’ is most often modified by postižený ‘impaired’ and poškozený ‘damaged, injured’ that create the extra negativity. Both adjectives were used in articles about adjusting the remuneration for disabled miners, a highly debated topic that year. Miners are once more in focus in 1999, but this time with the adjectival modifier zraněný ‘hurt, injured’. The news value increased that year due to an earthquake that had a severe impact on a mine near Ostrava. In addition, the dataset includes the adjective uvězněný ‘imprisoned’ or ‘trapped’, adding to the discourse of mining as a dangerous occupation. The noun horník ‘mineworker’ is the most commonly co-occurring noun with this adjective.

Two general nouns dominate the final ISCO 9 group: dělník ‘worker’, which mainly receives negative evaluations, and pomocník ‘helper, assistant’, that mainly receives positive evaluations. Of the 113,015 co-occurrences in the dataset, 51,528 include the former and 30,124 the latter. Kočí ‘coachman’ which is most often a surname, is a distant third, followed by uklízečka ‘female cleaner’ – one of very few female occupations in the most frequent co-occurrences. The negative co-occurrences are dominated by dělník ‘worker’, and in second place is pomocník ‘helper, assistant’. In this ISCO group of unqualified occupations, the noun dělník ‘worker’ is frequently modified by nekvalifikovaný ‘unqualified’.Footnote ²⁶ Other common modifiers include obyčejný ‘ordinary’ and ochotný ‘willing, co-operative’.

The positive co-occurrences are reversely dominated by pomocník ‘helper, assistant’ before dělník ‘worker’, and no other nouns enter the top five. When expanded to the top ten most frequent co-occurrences, the female worker dělnice squeezes in twice during the 1990s (the adjectives being krásný ‘beautiful’ and vhodný ‘suitable’;Footnote ²⁷ see Section 3.3). The reason we do not see the female cleaner here, despite uklízečka being a comparatively prominent noun, is that the distance is wider between the co-occurrences of noun and adjective – the adjusted sentiment value is not high enough for the female cleaners.

Summarising, the adjectives often describe the workers or their occupational hazards. They did not reflect any occupational pride nor positive community feeling; instead, the positivity often came from enthusiastic hobby representations or the presence of saints. As regards the theory on giving service to others leading to more frequent and more positive media representation, it cannot be confirmed here. Perhaps the lack of a supervisory function is the key to this, something that a future study could assess in greater detail. Finally, over time it became clear how the significance of farmers has changed after 1989, and the influence of climate change on the country become apparent through reports on their crops.

3.3.1 Noun Categories

The noun analysis reveals the proportions of nouns in each of five proposed categories (see what follows). This was carried out to enable comparison with previous research by Coffey-Glover (Reference Coffey-Glover2019) and to a certain extent also Havelková (Reference Havelková2017), Barát (Reference Barát and Lazar2005) and Ibroscheva and Raicheva-Stover (Reference Ibroscheva and Raicheva-Stover2009). It also considers the historical context of women being the core of the family (e.g. Feinberg Reference Feinberg, Wingfield and Bucur2006a,b), as described in Section 1.2.1. Mahlberg (Reference Mahlberg2005: 102) structured the people nouns in her study into similar categories, albeit based on meaning in her (English) corpus. In this section, the category ‘family’ is similar to her meaning 1, and includes generic nouns for men (muži), women (ženy) and family members. The five categories, partially informed by pre-existing variables for nationality and occupation, are ranked according to frequency, and frequency of modification by evaluative adjectives, in the dataset.

Consequently, one variable in the dataset comprises the subsequent five categories (listed here with four examples each): family members, FAM, e.g. žena ‘woman/wife’, babička ‘granny’, muž ‘man/husband’, dědeček ‘granddad’; humans in general, HUM, for example dáma ‘lady’, důchodkyně ‘female pensioner’, světák ‘man of the world’, důchodce ‘male pensioner’; nationalities, NAT, for example Norka ‘female Norwegian’, Syřanka ‘female Syrian’, Nor ‘male Norwegian’, Syřan ‘male Syrian’; occupational titles, PRO, for example soudkyně ‘female judge’, prodavačka ‘saleswoman’, soudce ‘male judge’, prodavač ‘salesman’; and religious affiliations, REL, for example Židovka ‘female Jew’, ateistka ‘female atheist’, Žid ‘male Jew’, ateista ‘male atheist’.

Table 7 shows the number of female and male nouns co-occurring with evaluative adjectives, aggregated over the whole period. The percentages in table 7 show that women are referred to as family members (FAM) in about the same proportion as men are referred to as professionals (PRO); in each case accounting for about 50 per cent of the female and male nouns, respectively. The second most common noun category for women in Czech media is their occupational role (just over 30 per cent). The men in this dataset have a more even distribution than women of other categories than their occupation (approximately 15, 11 and 21 per cent for FAM, HUM and NAT, respectively).

The most common nouns žena (43 per cent of the feminine nouns over the whole period) and muž (12 per cent of all the masculine nouns) are included in the category FAM, for family, following Mahlberg (Reference Mahlberg2005: 100). They mean both ‘woman’ and ‘wife’, both ‘man’ and ‘husband’, respectively. Both žena and muž account for just over 80 per cent of the FAM category co-occurrences for their respective genders. This is expected as these nouns have multifunctional uses and can also refer to biological differences in Czech. Even if half of the occurrences of these two nounsFootnote ³⁹ were in FAM and the other half in HUM, the ranking in Table 7 would remain unchanged. The exact proportions need to be investigated further. The term for ‘human’, člověk has been excluded from the sample in this section. It refers to any kind of human, despite its masculine grammatical gender (Čmejrková Reference Čmejrková, Hellinger and Bußmann2003: 33), and has the category HUM in the larger dataset.

Table 3 suggests that the men are predominantly referred to by their occupation (51.32 per cent) while the women are referred to as family members (50.83 per cent). The masculine nouns are also referred to using a wider perspective: the second highest category, NAT for nationalities, accounts for slightly over one-fifth, 21 per cent, of the total, two for 11–16 per cent and only one (religious affiliation) for less than 10 per cent. The nouns in the PRO category are analysed more closely in Section 2.4 and the NAT categories in Section 1.4. In contrast to men, the women maintain professional status (PRO) as their secondary most frequent category, making up nearly a third of the total feminine nouns. Within the dataset, the most frequently occurring PRO nouns among women include učitelka ‘teacher’, starostka ‘mayor’, pracovnice ‘female worker’, ministryně ‘minister’ and modelka ‘female model’. They are all in the top ten most frequent for women. For men, the top ten includes more occupations, of which the top five are starosta ‘mayor’, ministr ‘minister’, policista ‘police officer’, lékař ‘physician’ and pracovník ‘male worker’. Teachers are undoubtedly more often women than men in Czechia (up until 1999, see Weiner Reference Weiner2007: 34), and women are claimed to get a more positive representation in the Czech news when they dominate an occupation (Elmerot Reference Elmerot2017). While teachers still claim high prestige in Czech society (numbers 3 and 4 on the ‘Prestige list’ since 2011, see Tuček Reference Tuček2019 and Section 2.2), the occupations of policemen and physicians are more clearly linked to prestige and power than the occupations of teachers and models.

Figures 10 and 11 display the distribution of these noun categories over time in the dataset, co-occurring with evaluative adjectives. The ratio of the frequency of the categories remains fairly constant over time.

Figure 10 Noun categories per year for female nouns, relative to tokens in the sub-corpus.

Figure 11 Noun categories per year for male nouns, relative to tokens in the sub-corpus.

These two figures have a y-axis showing the number of occurrences relative to the total number of tokens in the sub-corpus used for this Element, to see changes in frequency over time. Having the relative number of tokens per year on the y-axis makes it a comparison with the whole population, to see whether the frequency of the noun categories increases or decreases compared to the whole population of news texts (see the corpus constitution in Section 1.3.4). In these cases, there are no overlaps between the evaluated noun categories.Footnote ⁴⁰ For the men, the occupational category (PRO) is the most prominent throughout the entire period (see Figure 11), and for the women, the most prominent is the family category (FAM), followed by the occupational category (PRO), (see Figure 11). We may, however, note the different scales for the genders, which shows that even the most frequent category for women in Figure 10, the family category (FAM), with frequencies at best between 150 and 210 per million tokens, would only be ranked third or fourth in relative frequency if placed alongside the masculine sample points of Figure 11. Nevertheless, the overall slightly upward trend for the female and the downward trend for the male co-occurrences are partly due to the evaluation as such and will be discussed further in Section 3.3.3.

Thus, in Figure 10, women are most often evaluated as a family member (FAM) over time; the trend line rises steadily after the initial year with few sources until its peak in the year 2000. In part, this might be explained by the contents of the corpus, where more media types are added that may have a larger interest in writing about women as well as men, such as certain lifestyle magazines. Additionally, the content of the constant news sources in the corpus changes; see Section 3.3.3. Occupational titles (PRO) form the second most frequent category. On the other hand, in Figure 11 men have a much wider gap between the category PRO and their second most frequent category, nationalities (NAT). The three categories NAT, HUM and FAM are closer to each other, and further from the PRO category than in the graph for feminine noun categories. This indicates that men are mentioned more frequently with their job title than anything else. The trend lines of these broad categories generally confirm previous studies (e.g. Zasina Reference Zasina2019; Elmerot Reference Elmerot2017). In Czech news, women are described from a relational perspective more often than men, but their occupational role is also important as a part of the whole picture.

3.3.2 Analyses Based on the Subjectivity Lexicon

This section provides an overview of the entire period using aggregated and partially separated data. Previous research rarely indicates whether women or men receive positive evaluations in the media, whether in Czech or other languages. Studies that do draw such conclusions mainly focus on a limited number of nouns for women and men, whereas this study offers a broader perspective.

The method used here stems from the premise that each year is a representative sample of the population of mainstream printed media of Czechia. As the number of sources grow, so does the corpus. The proportion of masculine and feminine co-occurrences within these years reflects linguistic changes in society and politics in that specific year. The summed sentiment is relativised in the graphs first by dividing it with the summed number of evaluated frequencies per issue and year (see Figure 12). Balancing out the amounts of gendered co-occurrences in this way helps to determine which gender is more positively evaluated. The sentiment values are then further normalised by the number of tokens per year in the corpus (see Figures 13–15).

Evaluation in the Whole Dataset

The increasing or decreasing frequency of evaluation is shown in Figure 12, using all the Subjectivity lexicon’s evaluative adjectives. This relates to the research questions by showing which groups – women or men – are more evaluated in the dataset, by positive and negative adjectives, respectively, and which trends are visible when looking at the evaluative aspect of language as a whole – thus combining the two facets. Figure 12 should be read as a balanced dataset, consisting of equal amounts of female and male co-occurrences, in an attempt to visually remove the androcentrism. It was created by dividing the summed adjusted sentiment value (see Section 1.3.5) with the total number of actual co-occurrences in this dataset. The evaluation is then presented in Figure 12 by positive (bright green) and negative (dark red) classification, respectively.

Figure 12 Proportion of positive (bright green) and negative (dark red) sentiment per year and gender in the dataset.

Figure 12 indicates that the women in my data are evaluated more in magazines and newspapers combined. Both the positive and negative facets reveal a larger proportion of female evaluation. However, we see that the positivity (bottom facet) decreases with time, whereas the negativity (top facet) only decreases slightly after 2001, then stays rather stable, and does not increase until after 2013. Therefore, the hypothesis that the women in this dataset are evaluated more than the men appears verified.

In contrast, the co-occurrences in Figure 13 are calculated relative to the total number of tokens in the corpus of newspapers and magazines (NMG) from which they were extracted. They reveal the amount of male versus female evaluated co-occurrences each year. Had there been an equal amount of feminine and masculine evaluated co-occurrences, the bars would have been equally long.

Figure 13 Relative frequency and sentiment classifications for the female (red) and male (blue) gendered co-occurrences per year.

Figure 13 reveals that masculine co-occurrence forms dominate throughout every year, indicating a high level of androcentrism. Additionally, it shows that women are both evaluated and mentioned more over time, occupying a larger proportion of the co-occurrences in later years. Combined, the two facets together demonstrate that the average Czech who has been reading newspapers and magazines since 1990 should become increasingly accustomed to women being (evaluatively) represented in the news.

Notably, there is a significant increase in positivity in 1995. Upon analysing the differences in adjectives between 1994, 1995 and 1996,Footnote ⁴¹ it appears they are not based on a specific event or a specific person, but on the corpus behind the dataset, where the lifestyle weekly magazine Vlasta was introduced in 1995 and then removed again. It would be expected that women were represented in a more positive light in a magazine like Vlasta.Footnote ⁴² The co-occurrences with the most positive impact from this magazine in 1995 include hodný + babička ‘nice + grandmother’, úspěšný + žena ‘successful + woman’ and skvělý + švadlena ‘excellent female dressmaker’. These co-occurrences appear frequently.

In summary, these quantitative analyses show that female nouns are slightly more likely to be accompanied by evaluative adjectives than male nouns. This lends support to the proposition that women are evaluated more than men. In addition, the proportion of positive evaluations for female nouns has been observed to decline over time, while the prevalence of negative evaluations has remained relatively constant until 2013, when it began to increase. Furthermore, the analyses demonstrate that masculine nouns are the most prevalent and that they exert a dominant influence on an annual basis, indicating a high degree of androcentrism. When considered together, the two analyses indicate that there has indeed been inequality in the evaluation of women and men in the Czech printed news media during these three decades.

3.3.3 Analyses Based on the Semantic Categorisations

Semantic categories are groups of words with a common denominator, which in turn gives the category its name. Some category examples are Evaluation: good/bad including words like atrocious or creditable, and Judgement of appearance including words like pretty or bleak. Since so few of the adjectives in the Subjectivity Lexicon were included in the classification made by Zasina (Reference Zasina2019), the semantic categorisation for Czech created at UCREL (see Piao et al., Reference Piao, Rayson and Archer2016) was used instead. More adjectives from this study were included there than in Zasina’s study, but 94 out of the 720 adjectives used with feminine nouns (and 98 of the 724 for men) had to be categorised manually for this study,Footnote ⁴³ following the UCREL Semantic Analysis System guide (Archer et al. Reference Archer, Wilson and Rayson2002) as closely as possible.Footnote ⁴⁴ The semantics of the adjectives is merely touched upon here, since the analytical perspective is intended to pave the way for future studies. It must be noted that some categorisations may not be immediately clear and require further discussion in more detailed studies.

Table 8 shows examples of adjectives from the Czech Subjectivity Lexicon, with the category/-ies to which it is assigned in the UCREL taxonomy. Each category is assigned a code consisting of a capital letter and a number,Footnote ⁴⁵ with or without sub-numbers. In addition, special characters are used, such as a + or – ‘to indicate a positive or negative position on a semantic scale’, and % or @ sign to mark rarity (see the whole list in Archer et al. Reference Archer, Wilson and Rayson2002: 1–2). As shown in Table 8, many adjectives are assigned to multiple categories in that system, complicating the analysis. The columns 1–11 are ordered by the ‘most likely tag first’ as described in Archer et al. (2002: 1).Footnote ⁴⁶ They thus indicate more likely to less likely semantic properties of the same adjective. For example, the adjectives těžký ‘heavy, serious’ and tvrdý ‘hard, stiff’ both belong to the category S1.2.5+ (Toughness; strong) and O4.1 (General appearance and physical properties). However, těžký is more likely to be semantically in group S1 than O4, whereas for tvrdý the opposite is true. Some more specific adjectives, such as pochybný ‘dubious, questionable’ and odrazující ‘discouraging’ only belong to one category each.Footnote ⁴⁷

To tackle the issue of multiple categorisations from above, only the categories listed in column 1 were considered in this section, as it is stated to include the most likely meaning for those with multiple categorisations. Figures 14 and 15 reveal the most frequent categories for each year. Excluding multifunctional adjectives such as velký ‘big, large’, dobrý ‘good’, poslední ‘last’, and rád ‘glad’ allows for a clearer understanding of the adjectives that make a difference.

Figure 14 Top (five including ties) semantic groups, excluding multi-purpose adjectives for feminine co-occurrences.

Figure 15 Top (five including ties) semantic groups, excluding multi-purpose adjectives, for masculine co-occurrences.

In Figure 14, five categorisations for the feminine co-occurrences remain consistent over time, with two consistently dominating.

In Figure 15, the decline for the masculine co-occurrences is consistent with Figure 13; men are evaluated less and less over time, and therefore appear less frequently in the dataset.

The following groups overlap, but the first is also the most prominent group of adjectives for women and the second for men here:

Some groups appear and disappear, such as G2.2 General ethics over time. For example, for the 1990s the feminine dataset includes svatý ‘holy’ and vystavený ‘exposed (to)’ or ‘displayed’ – a glance in the corpus indicates that they are mainly present in the literary supplement to the morning paper Lidové noviny that had a monthly circulation of around 80,000 copies during this decade.Footnote ⁴⁹ Again, the corpus was rather small in those years, hence the average value here gives even low frequencies a higher prominence than in later years. In 2004 and 2005, the adjective stereotypní ‘stereotypical’ boosts the semantic category A6.2 Comparing: usual/unusual as shown in Figure 14. It is important to note that even if adjectives co-occur with nouns, they may be far apart and only provide associations rather than actual collocations. For these semantic analyses, the distance between noun and adjective is weighed into the calculations, as in the subjectivity analyses in Section 3.3.2.

Compared with Coffey-Glover’s (Reference Coffey-Glover2019) table of semantic categorisation of masculine adjectives in women’s magazines from 2008 (112), the category ‘important’ is consistent with this Czech data. Her top findings – ‘shy’ and ‘romantic’ in the first place, and ‘dark’ and ‘stocky’ in second – are, on the other hand, not all present in my data, due to the short Subjectivity Lexicon: plachý ‘shy’ is there, but with less than 0.05 per cent of the total amount of adjectives for men (and just above 0.05 for women) during most years.

The overall conclusions from this calculation of frequency of each semantic group are that the categories denoting judgement of appearance, happy/sad and evaluation of good/bad are more prominent with feminine nouns, while adjectives modifying masculine nouns more often co-occur with adjectives denoting excellence and strength. The main category found both in previous research and here is ‘important’, which also in these data mainly co-occur with the masculine nouns.

3.3.4 Adjectives and Noun as Co-Occurrences

This section focuses on the individual adjectives co-occurring with the two genders of this section (see Section 1.3). The adjectives that most likely modify the nouns, namely those that co-occur within one word of the noun, will be examined first. This overview is comprehensible when comparing only two groups, whereas it would be overwhelming in Section 2 on occupations.

Firstly, the adjectives co-occurring with nouns referring to women will be discussed. Figure 16 shows the quantitative data for the whole period: the top five adjectives per year (no negated forms were as frequent), apart from the multifunctional ones (see Section 3.2). For all years, the adjective that defines women in this dataset is krásný ‘beautiful’. Every year, except for the first year when it only occurred on four occasions out of 239, this adjective is the most frequently used one to represent women in this dataset.

Figure 16 Top five (including ties) most frequent adjectives co-occurring with feminine nouns, per cent of female adjectives in the dataset per year.

In Figure 16, ‘beautiful’ first reaches a peak in 1992, and remains the most frequent adjective from 1997 to 2018, except for 1996 when it is overtaken by sociální ‘social’. This latter adjective is the second most frequent one over the years, as it can modify a variety of nouns, including ‘democrat’, ‘worker’ or even ‘prostitute’ (sociální pracovnice is a worker in both senses, although the latter meaning seems not to occur in the news). Additionally, the term appears in ministryně sociálních věcí ‘female minister of social affairs’. During the first years, the term was mainly used to refer to social workers, which may be a remnant of news values from the previous regime.

Only two adjectives remain constant throughout the period: zlý ‘bad, mean’ during the first years, although not present in the top list of adjectives used during these years. This adjective is almost exclusively used with the nouns žena ‘woman’ and čarodějnice ‘witch’ but also with sestra ‘sister’ during 2011–2013. Upon examination, the vast majority of co-occurrences over time are either related to the Shakespearean play Taming of the Shrew, which in Czech is Zkrocení zlé ženy, or to a book entitled Zlá žena ‘Evil Woman’. It does, however, also appear in sentences like Protože moudrá žena, která umí s mužem zacházet, dosáhne mnohem více než žena zlá a hysterická ‘Because a wise woman, who knows how to handle a man, achieves much more than a mean and hysterical woman.’Footnote ⁵⁰ The co-occurrence of zlý with sestra ‘sister’ or ‘nurse’ is partly due to a comparison with the Cinderella fairy tale that a lawyer made about the new constitutionFootnote ⁵¹ and partly due to a new play with two ‘evil sisters’.Footnote ⁵² The final constant is zraněný ‘hurt, injured’. It is primarily used with the noun žena ‘woman’ but can also modify důchodkyně ‘female pensioner’ and co-occurs in the vicinity of hospital nouns such as lékařka ‘female physician’ and sestra ‘nurse’. In other words, the adjective zlý, though frequent, rarely has the function of evaluating real-life women in the news, and zraněný often, but not always, occurs around female nouns of the caring professions handling hurt people. This serves as an example of the importance of handling large amounts of language data with care when drawing conclusions.

The discussion of individual adjectives is now reiterated, with the focus on those that co-occur with nouns referring to men. Figure 17 shows the diachronic frequencies.

Figure 17 Top five (including ties) most frequent adjectives co-occurring with masculine nouns, per cent of male adjectives in the dataset per year.

Many adjectives take turns to compete for a top three ranking annually. A consistently high-ranking adjective is zraněný ‘hurt, injured’ that appears in 1994. (It is worth noting that there were officially no criminals running around hurting people during the previous regime; see e.g. Štanzel Reference Štanzel2012:140). This term remains in the top three until 2018 and is mainly used to modify the nouns muž ‘man’ and mladík ‘young man’ followed by voják ‘soldier’ and policista ‘police officer’.

The second most frequent adjective of the period is the increasingly used bohatý, ‘rich’. This takes an extraordinary journey throughout the period from its debut in 1997 mainly referring to rich Americans and Russians but also Germans, Czechs and Jews (often in the context of either antisemitism or World War II). Despite a dip in 2009, the year after the European economic crisis of that decade, it peaks in 2014 and stays the most frequent adjective even after it has declined. In 2014, the nouns modified by bohatý are much more diverse than in 1997, and includes Ukrainians, Chinese, Indians, Jews, aristocrats and noblemen. From Section 3.3.1 it is clear that the semantic noun category REL contains a very small proportion of the nouns, but the use of bohatý significantly elevates the visibility of one noun (Jew) within this category. Related to ‘rich’ is slavný ‘famous’, which appeared within the top three rankings in 1999 with intermittent appearances until 2014, subsequently settling into third place. This often goes together with umělec ‘artist’ (see also Section 2.5.2 on the occupational group Craft and Related Trades), muž ‘man’ and bratr ‘brother’, and slightly less often with Americans, Czechs and a few other nationalities. Of the occupational titles, vědec ‘researcher’ is the most prominent of the modified nouns.

Another of the more constant, recurring adjectives, albeit not in the top three every year, is stínový ‘shadow’, mainly used about shadow ministers (of the parliament) or shadow city mayors. See also the introduction to Section 3 on the shadow government of women in 2000.

This study of individual adjectives shows that the adjectives co-occurring with nouns referring to women and men in the corpus exhibit distinct patterns. The adjective krásný ‘beautiful’ is the most frequently used to describe women. In contrast, the adjectives co-occurring with nouns referring to men are more varied, with zraněný ‘hurt, injured’ and bohatý ‘rich’ being consistently high-ranking and slavný ‘famous’ also appearing intermittently. Furthermore, the study emphasises the significance of contextual considerations in the use of adjectives, as evidenced by the co-occurrence of zlý ‘bad, mean’ with the noun žena ‘woman’. This co-occurrence is largely attributable to references to the Shakespearean play Taming of the Shrew.