Style and Rhetoric of Spanish Politics on Twitter

This article studies the communication strategies used in campaign messaging on Twitter by Spanish political parties during Spain’s 2019 General Elections in order to gauge whether a quantifiable relationship can be established between the style and rhetoric of a party’s Twitter speech, political platform, and political ideology. The analysis focuses on the discursive and rhetorical tactics that surround the parties’ engagement with issues of gender and feminism, particularly pressing concerns during this election cycle due to increased attention to gender-based violence and the organizing of feminist strikes in March 2019. In response to methodological questions surrounding the study of online speech, the study uses a combination of quantitative and qualitative methods to evaluate word choice, positive and negative sentiment, and use of platform infrastructure such as hashtags. Applying Natural Language Processing (NLP) techniques, the article examines word frequency, co-occurrence of qualified nouns, and sentiment analyses of tweets published by the five largest political parties in Spain between March 1 and May 15, 2019. Based on topic modelling, this corpus of tweets was then narrowed to those concerning gender and feminism and a close reading was conducted in order to locate the tweet’s ideological and discursive messaging within Spain’s sociopolitical context. Although word frequency analysis demonstrated that gender remained a concern for all five parties, noun co-occurrence and sentiment analysis revealed significant differences in how all parties engaged with gender as a political issue via their choices in rhetoric and style, which were linked to their platform and ideology via quantifiable measurements and qualitative close readings. As such, the study is able to conclude that using a combination of quantitative and qualitative methods enables researchers to draw nuanced and contextualized connections between the rhetoric and style of online political speech and the position of a political party on a given issue.

This article studies the communication strategies used in campaign messaging on Twitter by Spanish political parties during Spain's 2019 General Elections in order to gauge whether a quantifiable relationship can be established between the style and rhetoric of a party's Twitter speech, political platform, and political ideology. The analysis focuses on the discursive and rhetorical tactics that surround the parties' engagement with issues of gender and feminism, particularly pressing concerns during this election cycle due to increased attention to gender-based violence and the organizing of feminist strikes in March 2019. In response to methodological questions surrounding the study of online speech, the study uses a combination of quantitative and qualitative methods to evaluate word choice, positive and negative sentiment, and use of platform infrastructure such as hashtags. Applying Natural Language Processing (NLP) techniques, the article examines word frequency, co-occurrence of qualified nouns, and sentiment analyses of tweets published by the five largest political parties in Spain between March 1 and May 15, 2019. Based on topic modelling, this corpus of tweets was then narrowed to those concerning gender and feminism and a close reading was conducted in order to locate the tweet's ideological and discursive messaging within Spain's sociopolitical context. Although word frequency analysis demonstrated that gender remained a concern for all five parties, noun co-occurrence and sentiment analysis revealed significant differences in how all parties engaged with gender as a political issue via their choices in rhetoric and style, which were linked to their platform and ideology via quantifiable measurements and qualitative close readings. As such, the study is able to conclude that using a combination of quantitative and qualitative methods enables researchers to draw nuanced and contextualized connections between the rhetoric and style of online political speech and the position of a political party on a given issue.

Introduction
In Spain, the expanding role of social media-and Twitter in particular-in political activism and campaigning has been linked to the rise of the Los Indignados antiausterity movement that erupted in 2011 and the emergence of new political parties that have fractured the country's bipartisan democratic history (Ramos-Serrano, Fernández-Gómez, and Piñeda 2018, 127). This period also witnessed a de-centring of traditional media (television, print newspapers) and electoral programmes as primary sources of information for the electorate, as voters increasingly began to rely on social media. A recent study by Kennedy and Prat places Twitter among the top fifteen news sources in Spain (Kennedy and Prat 2019, 49). Voters who desire a more interactive and real-time discussion of political and electoral issues turn to Twitter-a "medium of immediacy" (Johnson 2012, 57)-to see how parties respond continuously to the flow of current events, engage their opponents, and speak directly to their audiences on a daily basis. Not only does the shift from television, newspaper, and electoral programmes to tweets represent a shift in medium, it also entails a change in parties' rhetorical and communications strategies and the electorate's reception of campaign messaging that urges scholarly consideration. As Janet Johnson argues, in exchange for widespread visibility and greater campaign reach, politicians must learn to adapt their rhetoric to the digital stage, especially on platforms like Twitter that require concise presentation of ideas (Johnson 2012, 55). Given this relatively new form of political campaigning and political engagement, the questions of how Spanish parties address their constituents within Twitter-mediated elections and what methodological approaches are best suited to analyze parties' campaign messaging online merit attention.
In response to these questions and in order to understand the relationships and differences between each party's rhetoric and style of tweets, their real-time responses to current events, and ideological stance on key electoral issues, this article employs various Natural Language Processing (NLP) techniques-namely word and hashtag frequency and sentiment analyses-to analyze tweets posted by the five most popular Spanish political parties in 2019-the Partido Socialista Obrero Español (Spanish Socialist Workers' Party, shorthand PSOE), Partido Popular (People's Party, shorthand PP), Ciudadanos (Citizens, shorthand Cs), Podemos (meaning "We Can"), and Vox-before, during, and in the immediate aftermath of the April 28, 2019, General Elections.
Previous studies of Spanish political Twitter discourse indicate that parties' adaptive strategies for digital communications have varied and tend to find that parties do not always take advantage of the interactive and communicative capacities of Twitter (see Zamora-Medina and Zurutuza-Muñoz 2014;López-García 2016;and Ramos-Serrano, Fernández-Gómez, and Piñeda 2018). However, these studies have stopped short of examining the stylistic and rhetorical strategies that parties use to communicate their positions digitally through the stream of abbreviated messages on Twitter. In their study of European political party campaigns, Crabtree et al. (2018) note, "[w]hereas campaign content and campaign focus address what parties say and who they say it about, campaign sentiment addresses how they say it" (Crabtree et al. 2018, 1). A focus on rhetoric and style enables this study to address the inter-related nature of political language and affective sentiment and the ways in which they are used together to inform and persuade potential voters. As such, our analytical framework aims to examine how political rhetoric online functions as a discursive tactic "to produce action or change in the world" (Bitzer 1968, 4). This study builds on existing scholarship by proposing a combination of quantitative and qualitative methods to accurately parse and analyze political Twitter speech, heeding the call to investigate "how they say it" by evaluating the words parties use, the presence of positive and negative sentiment in their tweets, the use of platform affordances such as hashtags, and close readings to situate parties' rhetorical strategies within their sociopolitical contexts.
To that end, this article begins with quantitative assessments-word frequency, co-occurrence (of qualified nouns), and sentiment analyses-of tweets published between March 1 and May 15, 2019, to gain a broad-level understanding of each party's preoccupation with key electoral issues and rhetorical approaches to addressing said issues on Twitter. Grounded in the results of these distant readings of each party's tweets, this analysis then narrows its focus by targeting and parsing tweets that contain frequently occurring terms within a specific topic category-in this case, gender.
Gender provides an ideal frame for quantifying stylistic and rhetorical differences between parties given that the collection period for this study coincides with significant political discussion about gender-related issues in Spain.
An example of such was the renewed media attention in spring 2019 to a court case revolving around the gang rape of an 18-year-old woman during the San Fermín celebrations in Pamplona, Navarre, in 2016. The two legal categories which address sex without consent in Spain are "sexual abuse" and "sexual aggression," the latter necessitating violence or intimidation, the parameters of which are extremely narrow and subject to interpretation. The perpetrators had been previously convicted of the lesser crime of sexual abuse instead of sexual aggression in 2018. This sentence exposed the failings of the Spanish legal system in addressing such cases and galvanized and enraged feminist groups. During the collection period of this study, the Spanish Supreme Court was deliberating a revised sentencing of these men. The perpetrators were finally reconvicted of "continuous sexual assault" in June 2019. This period also spans the #8M feminist strikes in Spain on March 8, 2019, in which gender and feminism were much-discussed topics in the cultural milieu, and points of conflict over which parties hashed out ideological and discursive conflicts.
During the #8M strike, an estimated 5 million Spanish women did not go to work or do any paid or unpaid labour to protest the wage gap. The strikers protested economic disparities, but also gender discrimination and sexualized violence, among other intersecting experiences of gendered oppression. Accordingly, our methodology also understands feminism and women's rights to be intersectional issues that cannot be separated from other key political topics during the election cycle, including employment, immigration, and health care. In fact, gendered social relations structure the way these other economic and social issues play out in a material way, impacting the lives of all Spaniards. Focusing on gender enables this study to examine both how political parties frame questions of gender and feminism as well as how they situate gender in relation to other political concerns.
By combining quantitative and qualitative approaches, as opposed to one or the other, to analyzing the topical focus and affective tendencies of each party's tweets, this study unearths a nuanced relationship between a party's discursive strategies on Twitter and their position on and ways of engaging with key electoral issues and inter-party disputes that is grounded in both a distant and close reading of relevant tweets. Rhetorical differences between parties are most pronounced in topical contexts, such as gender-related discussions, where politicians resort to a variety of discursive mechanisms, such as word choice, sentiment, and hashtags, to distinguish their viewpoints from those of their political opponents. Using the Spanish General Elections of April 2019 as a case study, this article provides a model of how to leverage each of these methods to maximize one's critical understanding of political Twitter and an in-depth study of the Twitter style and rhetorical strategies of Spain's established and emergent parties when engaging with key and often contentious electoral issues.

Background
In the last five years, Spaniards have witnessed the fracturing of the historical political dominance of the PSOE and the PP, the two parties that held mostly uncontested control over the country since the 1977 elections that followed the end of the Franco regime.

Data collection and methodology
Between March 1 and May 15, 2019, we collected 10,038 tweets posted by the Twitter accounts of the five major Spanish political parties: @PSOE, @Populares, @ CiudadanosCs, @ahorapodemos, and @vox_es (counts are listed by party in Table 1).
Twitter's Application Programming Interface (API) offers both streaming and search services. The former allows for real-time monitoring of tweets based on predefined search criteria (e.g., account names, hashtags, keywords, or geographical areas). For this particular study, the streaming service was used to target and monitor discussions related to the 2019 General Elections. The latter search service facilitated the gathering of profile and timeline information from the five accounts. Collected data was then saved in a NoSQL database to streamline queries and future analysis, consisting primarily of NLP techniques for word and hashtag frequency analysis, the creation of co-occurrence networks of nouns and their qualifiers, and sentiment analysis. This database was also used to withdraw tweets for selected close readings. This study uses the Stanford CoreNLP tool given that it includes a robust Spanish language model, with an accuracy rate of 97% per token (words and punctuation symbols) and at least 55% per sentence (i.e., if only one token in a sentence is incorrectly classified, the whole sentence is marked as incorrectly tagged) (Manning 2011). NLP enables the extraction of the full sentences embedded in tweets as well as the words in each sentence. Single words and words within sentences are be tagged based on their parts-of-speech (POS) classification (nouns, qualitative nouns, verbs, and adjectives), while taking typographical and rhetorical devices such as punctuation into consideration. These classified terms can then be processed and visualized for further quantitative and qualitative examination. As this study aims to first provide a quantitative analysis of the rhetorical strategies used in tweets published by political parties, NPL techniques provide a number of methods to do so, including Latent Dirichlet Allocation for topic modelling, absolute and relative word frequency for general content analysis, and sentiment analysis for numerical measurements of positive and negative messaging within a given tweet.
A closer look at the parts of speech used most frequently by each political party in their tweets enabled us to identify central party concerns (nouns were especially indicative of these), recognize stylistic and rhetorical strategies (word choice and use of hashtags), and explore relationships between the speech, ideology, and campaign platforms of the leading political parties. After conducting an initial word frequency analysis, we created co-occurrence networks to examine how parties qualified nouns related to gender (mujer, mujeres, and feminismo, meaning woman, women, and feminism).
Co-occurrence networks rely on matrices that organize selected words in relation to the context of their sentence. For this project, a set of pre-and manually selected target terms related to gender were used with the goal of understanding how different parties address the issue of women's rights and how they view feminist movements.
In addition to parsing tweets for frequently used terms as a means of identifying key electoral issues, we conducted sentiment analysis to gauge the tone (positive, negative, or neutral) of the messages shared by each party. Sentiment analysis tools offer a means of processing relatively large amounts of text to reveal affective and/or emotive tendencies. Typically, this approach relies on the polarity of a positive/neutral/ negative classification achieved through the application of lexicon-based approaches.
For example, generally positive words, such as nice, good, and fabulous, are assigned Obs.: The figure in parentheses denotes the total number of original tweets originating from each party's account, excluding retweets. On March 8, 2019, International Women's Day, the PP changed its official Twitter handle from @ppopular to @populares. This study considers the data from both accounts as a single data point.
higher sentiment scores, while commonly negative words, such as terrible, bad, and corrupt are given lower sentiment scores. Negation words, such as not, but, and however, are also taken into consideration when sentences are evaluated for sentiment. For this study, we chose Google's Natural Language API, as it supports sentiment analysis in Spanish. As is typical for most sentiment analysis tools, Google's API normalizes the sentiment score of a sentence within the range of -1.0 (negative) to +1.0 (positive).
The first step of the sentiment analysis was an assessment of all tweets posted from each party account. We then ran a more nuanced sentiment analysis of tweets containing key gender-related terms-specifically, mujer (woman), mujeres (women), feminismo (feminism), and feminista (feminist)-to better understand the relationship between differences in party sentiment and the rhetorical treatment of gender equality and feminism. Despite notable advances in sentiment analysis research in recent years, the complexity and nuances of human language make the task of sentiment analysis nonetheless challenging. With that said, sentiment analysis is useful when engaged as one of several analytical tools with which to assess political speech. To illustrate stylistic and rhetorical differences between parties, this study begins with word frequency analysis to identify topics of interest, constructs co-occurrence networks to examine the use of selected terms in context, and then turns to sentiment analysis to look at the emotive dimensions of party tweets. Each of these processes is accompanied by close readings that situate the quantitative results in context and elaborate on subtle differences in distant readings that point to significant variations in party rhetoric. The final portion of this article uses a word frequency analysis to identify frequently used hashtags to examine the use of these hashtags as discursive frames.

Ethics
While the content of this study involves cultural conversations around sexualized violence, we chose to focus exclusively on the tweets of official party accounts, with no reference to the tweets of individual users. Many users were engaging with several of the hashtags through self-disclosure of experiences of sexualized violence and use of this information in a study would necessitate careful and prolonged ethical consideration.
While the generally accepted best practice in ethical research involves verifying if tweets remain public at the time of publication, there is ongoing discussion in digitally engaged fields about how to navigate the complexities of online data harvesting. Jackson, Bailey, and Welles suggest that writing about individual users' tweets, particularly "marginalized people who are not in the public eye … risks exposing them to unwanted and unanticipated attention" (Jackson, Bailey, and Welles 2020, xi). In their 2020 book, Jackson et al. detail the additional precautions they employed in the data collection process to address the tension between representing Twitter conversations and discourses accurately and honouring users' right to privacy and ownership of their own data. For instance, they exclusively selected tweets that were not only public, but addressed to a larger community, excluding replies on individual user's threads. To respect privacy, they also disqualified tweets from closed accounts, as well as tweets which had been subsequently deleted. Other scholars, such as Earhart, underline the importance of not abstracting data from its humanity, as it is "always a part of a community or individual" (Earhart 2018, 369). She asserts that without attention to issues of consent, ownership/control of data, and without situating oneself as a researcher, we can easily exploit communities with which we engage (Earhart 2018).
To ethically include individual user's data, especially that which could be raw or personal, this study would have been designed differently. The goal of this work was to investigate political parties' emerging strategies of digital communication, and we believe that within the scope of the project, we could not dedicate the deserved amount of proper attention and care to individual user's data, and thus it is excluded from the study.

Overall word use
We structured our data to determine which terms appeared most frequently in party tweets and selected the top twenty nouns used by each party based on the assumption that nouns are better indicators of party issues and concerns than verbs and adjectives.
We excluded proper nouns with the exception of España (Spain), the names of the five party leaders (Pedro Sánchez, Pablo Casado, Pablo Iglesias, Santiago Abascal, Albert Rivera), and the names of Spain's autonomous regions and cities, and provincial capitals.   priorities, but even when party concerns demonstrate some overlap, a party's word often reveals an implicit ideological stance on an issue. This demonstrates how parsing the language of tweets provides readers with insight into the relationship between a party's ideology and word choice in Twitter campaign discourse.
Repeat occurrences of nouns across party accounts paint a picture of general electoral concerns, highlighting matters that all parties address in their tweets. For instance, España (Spain) appears within the top noun count for all five parties (see Table 2). This is to be expected given that these are national elections. Other topics (nouns) discussed by all parties include mujer (woman)-also unsurprising given that these elections take place shortly after the Women's Strike on March 8-and voto (vote), although how each party engages with these matters varies. However, beyond identifying common electoral concerns, this study finds that a comparison of noun recurrence between parties also reveals differences in each party's electoral priorities and, at times, suggests their stance on an issue.
In the case of Catalan neo-liberal party Ciudadanos, there is a high occurrence of terms associated with Catalonia (Cataluña, n = 88, 1.0%) and freedom (libertad, n = 188, 2.2%), reflecting the party's neoliberal stance and opposition to Catalan separatism, in favour of a pro-Spanish, pro-European, and post-nationalist stance. In that of populist and self-proclaimed feminist party Podemos, the terms people (gente, n = 271, 3.7%), woman (mujer, n = 75, 1.0%), and feminism (feminismo, n = 47, 0.7%) recur often, highlighting its interest in discussing and appealing to "the people," women, and feminists. As a point of contrast, the fiscally conservative right-of-centre Partido Popular discusses taxes (impuestos, n = 160, 1.3%) more often than any other party, as well as government (gobierno, n = 361, 3%) and the presidency (presidente, n = 202, 1.7%) more than any other topic besides Spain (España, n = 529, 4.4%). PSOE's second-most-tweeted term after Spain (España, n = 463, 5.5%) is right (derecha, n = 313, 3.7%), followed by government (gobierno, n = 237, 2.8%) and woman (mujer, n = 202, 2.4%). This suggests that the party's main electoral priorities include opposing the rise of arguably the biggest threat to its power, what it refers to as the right-wing trio (trío de derechas)-comprised of PP, Ciudadanos, and Vox-and women's empowerment, rights, and freedoms. Finally, Vox extensively tweets the terms rally (acto, n = 176, 3%), a reference to its many campaign rallies, and list (lista, n = 101, 1.7%), a reference to the electoral lists of provincial congressional candidates. A relative newcomer to the political scene in Spain, Vox frequently uses nouns that demonstrate the party's interest in self-promotion, generating a sense of community through numerous rallies, and introducing its provincial candidates to its voter base online. This contrasts with the incumbent leftist PSOE's more defensive and oppositional stance towards the right, as it focuses on maintaining and defending its power and popularity, rather than trying to obtain them like Vox.

Word use and gender
While the recurrence of a noun-or electoral issue-suggests the degree to which a party is preoccupied with that topic, frequency of word use is not necessarily indicative of a party's support for that issue. Our gender case study illustrates this point. For instance, neoliberal party Ciudadanos tweeted the term feminismo (feminism, n = 52, 0.6%) approximately as often as feminist parties Podemos (n = 47, 0.65%) and PSOE (n = 44, 0.52%). However, few are likely to argue that Ciudadanos is equally or more feminist than Podemos. If anything, Ciudadanos is criticized by feminists, journalists, and leftleaning political pundits for its neoliberal pseudofeminist discourse that envisions a feminismo liberal (liberal feminism), which advocates for women having the freedom and right to make decisions over their own bodies without proposing policy changes that will actually empower women, in the same vein that it argues for neoliberal, freemarket, and capitalist economic policies. In the words of feminist activist, journalist, and author Cristina Fallarás, "El feminismo tiene que ser anticapitalista por definición porque el capitalismo es patriarchal" ("feminism must be anti-capitalist by definition because capitalism is patriarchal," translation mine, quoted in Trobat 2019). So, although feminism is a recurrent theme in Ciudadanos's tweets-highlighting their consistent concern and engagement with the issue-their position on feminism is not captured in a basic word noun count or word frequency analysis. To see if these discursive differences in the contextual use of gender-related words can be made on the macro level through quantitative analysis, we conducted a co-occurrence analysis to identify the qualifiers surrounding mujer/es (women) and feminismo (feminism).

Qualified nouns
To develop broad-level assessments of parties' positions on the recurring issues identified in the word-frequency analysis, this study delves further into the gender case study by using co-occurrence networks to compare parties' use of qualified nouns. While the word frequency analysis quantifies the extent to which a particular noun/issue/topic is discussed by a party, the study of qualified nouns provides an indication of how and in relation to what those issues are discussed. From the results of the word frequency analysis, we selected the terms mujer/mujeres and feminismo to further nuance our gender case study. The node graphs in this section (Figures 3, 4) represent the results of this analysis: first, the qualifiers that each party associates with the term in question and, second, the relative usage or recurrence of each qualifier in tweets. The latter is represented by the relative width of the lines, or edges, connecting the central term to its qualifiers, with thicker and thinner edges representing higher and lower frequencies of use, respectively. Qualifiers need to have occurred a minimum of ten times to be selected to the graph.
As Figure 3 reveals, the PSOE most often qualifies mujer and mujeres (woman and women) as joven (young), española (Spanish), and asesinada (murdered). Podemos, meanwhile, qualifies mujer/es in terms of a group or gathering (concentración de), a discursive tendency in line with the party's focus on strike-and demonstration-related frames for discussing gender-related issues (see further discussion of this tendency in Hashtags section). In contrast, the PP qualifies mujer/es as trabajadora (working) and embarazada (pregnant), reflecting the party's focus on workplace and family policies when addressing gender. Echoing the party's focus on liberalism, our analysis of Ciudadanos's tweets containing mujer/es yields the qualifiers valiente (brave) and autónoma (autonomous/independent). In contrast, Vox qualifies mujer/es as española (Spanish) and maltratada (abused or mistreated). While these qualifications shed some light on how each party views and represents women and their issues and engages with issues such as gender violence, perhaps most telling is that all five write about women in relation to men, hence the prominence of hombre y (man and) as a prominent qualifier for mujer/es. In fact, except for Podemos, all parties qualify women in relation to men more so than anything else, indicating that they consistently imagine women within a male-female binary-rather than as independent from men-and, perhaps, that they wish to assure voters that the party sees a place for men within their discussions of women's rights and feminism.
Moving from mujer/es to feminismo-a term that sparked substantial debate among parties-again reveals differences across parties, shown in but those that do tend to deflect critique of the party's policies in favour of levelling criticism at their left-leaning detractors on the grounds that these entities promote an exclusive and/or negative form of feminism. Tweets containing these qualifiers yield messages in which the PP's politicians protest the perceived negativity of the party's critics, rebuffing the lección sobre (lesson on/lecture about) feminism offered by those who criticized the PP's stance, on the grounds that the PP already espouses a feminismo positivo (positive feminism) that positions women as equal to, rather than against, men.
No hay que confundir la reivindicación del #8M y aquello que quiere utilizar la izquierda en la manifestación, que es la idea de que el feminismo solo puede ser de izquierdas. A la mujer no hay que imponerle un dogma para ser feminista. However, feminismo liberal is also the most frequently tweeted qualifier-noun pairing of the PSOE. According to this node chart, then, the PSOE and Ciudadanos most often qualify feminism in the same way, as liberal. This similarity, however, is not a good indicator of each party's stance on feminism, given that, while Cs promotes a liberal feminism, PSOE does not. Determining the context in which a qualified noun is tweeted demands a close reading of tweets containing the term feminismo liberal. In fact, PSOE's tweets containing feminismo liberal reveal that the qualified noun is only mentioned in tweets questioning the notion of liberal feminism: Queremos un país con mujeres libres, seguras y vivas. La derecha habla de feminismo liberal y violencia intrafamiliar. Las palabras en democracia representan compromisos y el PSOE será el valladar de los derechos y libertades de las mujeres.
(We want a country whose women are free, safe, and alive. The right talks about liberal feminism and inter-family violence. Words, in a democracy, represent commitments, and the PSOE will be the defender of the rights and liberties of women.)

(PSOE [@PSOE] 2019b)
In this light, the PSOE's other frequently used qualifiers-nuevo modelo de feminismo (new model of feminism), and feminismo en estado puro (feminismo in its pure state)seem to be a defense of endorsing feminism unconditionally. While Podemos also engages with Ciudadanos's platform via the use of liberal as a qualifier, their qualifiers illustrate the party's call for a "plural" feminism that is intersectional and diverse, one that fills the streets (calles), echoing their use of concentración de to qualify mujeres.
Notably, Podemos's call for policy changes that would include a mandatory asignatura de feminismo (course on feminism) being taught in Spanish public schools drew the ire of Vox (see the above discussion of Vox's use of cursos de). Podemos calls for a feminist revolution, the creation of feminist policies and a strongly worded platform that advocates systemic overhaul, mobilization, and change, whereas Vox attempts to discredit feminist movements, parties, and organizations by referring to them as supremacista (supremacist) and imperante (dominant).
The quantitative analysis and visualization of qualified nouns alone, whilst signalling how key terms like feminism are most often qualified in party tweets, fall short-like the noun frequency analysis ( Table 2)

Sentiment analysis: Assessing the affective dimensions of tweets
Political strategists and campaign managers have long realized that emotion is key to mobilizing voters and garnering their support. Considering this, this study analyzes all collected tweets to obtain a macro-level analysis of party variations in sentiment. This analysis also examines specifically gender-related tweets to assess the relationship between party stance on key campaign issues and sentiment to better understand the affective nature of a party's rhetoric and campaign tactics on Twitter as it relates to gender and feminism. Using Google's sentiment analysis tool, sentiment was evaluated on the level of each tweet, regardless of the length of the tweet. are the same parties that express more positive sentiment in their tweets, suggesting they have a higher degree of satisfaction with Spain's current political system, whereas less mature parties display more negative sentiment in tweets, suggesting their lower degree of overall satisfaction with the system. A close reading of the following tweet from the PP offers a more qualitative illustration of these rhetorical tendencies. In making their promises to the electorate, the incumbent parties PP and PSOE strive to portray their political records as works-in-progress that should be interpreted in a generally favourable light. However, these differences in positioning do not translate to stark differences in sentiment analysis scores between the parties: opposition parties Podemos and Vox have only slightly lower median sentiment scores (0, indicating overall neutral sentiment) compared to the incumbent parties and Ciudadanos (0.1).

Comparing overall sentiment
To better leverage the capacity of sentiment analysis to draw broad-scale conclusions about party rhetoric, it is helpful to compare how party sentiment in discussions of a particular topic-gender-compares with the sentiment scores of their tweets overall.
This comparative approach offers a more nuanced reading of party differences and allows us to identify areas for further selected close readings.

Sentiment analysis using gender-related terms
For a more nuanced understanding of each party's stance regarding topics related to gender, we performed a sentiment analysis of tweets containing the frequently occurring nouns mujeres, feminismo, and feminista identified in the word frequency analysis. This analysis was done by looking at each party's tweets that included the gender-related target terms feminism and feminist. We created a corpus of tweets containing the terms in question, and then used Google's Natural Language API to compute the individual score of each sentence from the filtered tweets. This analysis enables us to form conclusions about the differences in parties' use of relatively neutral gender-related terms, like mujeres, versus contentious terms that attracted political controversy during the leadup to the elections, like feminismo (see Figure 6).
Our comparative study of mujer/es and feminismo/feminista indicates that the five parties tended to treat these terms distinctly. Notably, with the exception of the PSOE, the parties' gender-related tweets tended to display more variable sentiment (as measured by the Interquartile Range, or IQR), indicating less consistency in their treatment of gender. Among more conservative parties (PP, Ciudadanos, Vox) tweets containing feminismo/feminista displayed more negative sentiment than both the parties' tweets overall and those containing mujer/es. This is especially interesting in the case of Ciudadanos, whose feminismo liberal platform drew substantial criticism from more progressive parties. In fact, the median sentiment score for tweets that Figure 6: A comparison of party sentiment based on three analyzed corpuses: all tweets, tweets that include mujer and mujeres (woman/women), and tweets that include feminista/feminismo. Note that the dots represent outliers.
included the terms mujer and mujeres was lower than the overall median sentiment score for all parties except Ciudadanos (whose median scores were the same across all three sentiment analyses). Figure 6 visualizes the sentiment of tweets that include feminismo and feministas from all five parties using box plots.
Although the PSOE, Podemos, and Ciudadanos share a median positive sentiment score of 0.1 in the feminismo/feminista sentiment analysis, there are pronounced differences in the sentiment patterns of these parties. The generally positive sentiment of Podemos' tweets containing feminist terms is unsurprising in that Podemos positions itself as a feminist party. Although Podemos is the only party whose sentiment surrounding the terms feminismo/feminista is more positive than that of tweets containing mujer/es and the party's overall sentiment, Podemos also displays a fair amount of negativity in these tweets. Upon closer examination, it becomes clear that Podemos adopts a strongly negative tone when criticizing the policies of other parties: ¿Cómo un partido que critica las subvenciones para la lucha contra la violencia machista puede llamarse feminista? Este es el #FeminismoLiberal de Cs : a favor de los vientres de alquiler y en contra del lenguaje inclusivo. Mucho tienen que aprender del movimiento feminista.
(How can a party call itself feminist and criticize policy changes to combat gender- Vox's sentiment around feminism and feminists is by far the most negative of the five parties, with the Partido Popular a near second. Upon closer inspection, however, the target of this negative sentiment was revealed to be substantially different between the parties. Although the overall sentiment of Vox's tweets leaned negative, their tweets containing the terms feminismo/feminista were especially so, with a median sentiment score of -0.2. This is especially significant given that Vox's sentiment in tweets containing the terms mujer/mujeres and feminista/feminismo is markedly more negative than the party's overall sentiment (for all tweets). A close reading of Vox's tweets that address the 8M feminist demonstrations and the notion of a feminismo supremacista further illustrates how this negative sentiment plays into an underlying assertion of Vox's platform.
Las mujeres de la España real no necesitan un colectivo que les diga lo que son, mujeres fuertes, trabajadoras e independientes. Ni un euro público más para los lobbies del feminismo supremacista que propagan odio y división.
(The women of the true Spain do not need a collective to tell them who they are, strong, hardworking, independent women. Not one euro more for the supremacist feminist lobbies that promote hate and divisiveness.) (VOX [@vox_es] 2019b) Here, by using sentiment analysis to guide close readings of the tweets, we identify points of discursive and ideological conflict between parties, particularly over the meaning of feminism and the role of the political party in advancing a specific interpretation of gender equality.

Hashtag analysis: Evaluating parties' framing of topics
Unlike simple noun use, hashtagging creates opportunities for discursive participation online and, to borrow from Benedict Anderson (1983), imagined communities of users who perceive themselves as part of a particular group or movement such as Los Indignados-#15M (Anderson 1983). Hashtags enable political discussion on Twitter to serve a social and dialogic purpose, linking users together based on a similar topic of interest. Although hashtags originated as a means of identifying and searching for discussion topics, Ash Evans (2016) argues that "the use of this affordance has shifted to become its own interactional communication" (Evans 2016), allowing users to create a dialogue and take a clear stance within the space of a single tweet. Similarly, Pond and Lewis (2019) identify hashtags as tools for the social production of meaning, as "genre defining discourses" through which Twitter users engage action frames (Pond and Lewis 2019, 217). The hashtag, along with the text of the tweet and any linked media, are components of a system of symbolic exchange through which Twitter users encode and interpret meaning.
In the context of this study, this means that while some hashtags function as labels that aggregate tweets based on a topic, like #España, others transmit campaign positions, like #MásPPMásIgualdad (#MorePPMoreEquality), or #CsConLas-Familias (#CsStandsWithFamilies) or prescribe an audience viewpoint, like #LaEspañaQueQuieresEsFeminista (#TheSpainYouWantIsFeminist). Analyzing topical hashtag use across parties reveals how they frame and engage with a given topic. By comparing the topical focus of the most-used hashtags of each party and further nuancing this analysis with a case study of gender-related hashtags, this section examines how each party strives to frame the discourses they engage on Twitter through strategic hashtag use.
Parties' engagement with gender-related hashtags situates them in relation to larger feminist movements and their goals and interventions.

Feminist and activist hashtag politics
Significant scholarship has been produced on feminist and activist hashtag use. For example, work exists on how the use of hashtags by feminist movements fosters online conversations on specific issues, such as misogyny in Korea (Kim 2017), or on hashtags as a form of feminist "shouting back" (Turley and Fisher 2018). Scholars have also cited media hashtagging as a response to rape culture, and as a method to "call out" and demand accountability from perpetrators "when mainstream news media, police, and school authorities do not" (Rentschler 2014, 67).
It has also been argued that hashtagging has brought ordinary people into the political arena and aided in feminist consciousness-raising on a large scale. Mendes, Ringrose, and Keller study how the hashtags #BeenRapedNeverReported and #MeToo do the work of creating feminist solidarity by exposing the structural nature of oppressive experiences, and concurrently make survivors feel heard (Mendes, Ringrose, and Keller 2019). This is also due to the conversational nature of the platform, which enables a larger "collective storytelling" (Jackson, Bailey, and Foucault Welles 2020). Larrondo (Larrondo, Morales-i-Gras, and Orbegozo-Terradillos 2019) similarly writes about this phenomenon in Spain, contending that hashtag activism "would appear to be an intermediate step in a longer process of creating a higher consciousness regarding gender equality issues in Spain" (207).
However, online feminist networks that utilize hashtagging are not without critique.
They have been largely problematized for replicating pre-existing racial hierarchies (Feldman and De Kosnik 2019). This results in a dominant digital politic that ignores the contributions of women of colour in activist movements and centres the experiences of white women (Mueller et al. 2021).
The feminist Twitter protest tactic of self-disclosure to gain visibility has also been discussed in terms of its potential harms. While it has certainly had deep social impacts, this visibility can simultaneously put survivors at risk. As Clark-Parsons writes, "publicly performing the identity of a survivor in a cultural context where sexual violence victims are shamed and doubted" (Clark-Parsons 2019, 10) can make participants more vulnerable to online harassment, doxxing, and personal attacks.
Parties' framing of political discourse is in conversation, reaction, and response to activist and feminist hashtagging practices, particularly in terms of gender issues.

Overall hashtag use
The results of our preliminary hashtag analysis are summarized in Tables 3 and 4. A superficial reading of these hashtags supports the broad policy leanings associated with each party. However, a look at the context in which each party's most frequently tweeted hashtags were produced reveals more complex entanglements of party selfrepresentation and use of hashtags to carefully frame political discourse.
The left-of-centre incumbent PSOE's use of #LaEspañaQueQuieres (#TheSpain-YouWant) reinforces the generally positive sentiment that the party displayed in this study's Sentiment Analysis. By linking this slogan with a demand for gender equality in #LaEspañaQueQuieresEsFeminista, the party pitches itself as not only a party with a proposed platform on gender equality, but a party that advocates for a feminist Spain and expects its voters to want the same. Notably, #LaEspañaQueQuieres and other PSOE tags like #HazQuePase (#MakeItHappen) and #VotaPSOE (#VotePSOE) directly address the audience with action-and future-oriented frames that invite the reader to be part of the PSOE's political project.

#MujeresFuertes
(1) (#LiveSpain) and #PorEspaña (#ForSpain), but also draws on anti-feminist rhetoric with #NoHablesEnMiNombre, which is discussed further in the next section. (#NothingCanStopUs8M) reinforce the assessment of Podemos's approach to feminism that emerged from the co-occurrence analysis-one that is based on a vision of feminism as an unstoppable wave that fills the streets.

Gender-related hashtags
While the more progressive parties tended to describe the 8M protests as a #HuelgaFeminista (#FeministStrike), conservative parties stuck to a muted playbook of celebrating women in family-and workforce-oriented frames. The conservative PP-who declined to participate in the 8M demonstrations on the grounds that the left-leaning parties had "politicized" the event-opted for variations of the tag #DíaDeLaMujer (#WomensDay) rather than invoke frames associated with feminist demonstrations. The centre-right and conservative parties-PP and Ciudadanostended towards hashtags that focused more heavily on loosely defined equality frames in the context of workplace and family policy, as the PP's #MásPPMásFamilia (#MorePPMoreFamily) and #LeyDeMaternidad (#MaternityLaw) indicate. The PP's use of #NoHablamosHacemos (#WeDon'tTalkWeGetThingsDone) is especially compelling, as it attempts to deflect the critique from left-leaning parties of the right's language surrounding gender by pointing to the PP's historical encouragement of women's labour force participation. Similarly, Ciudadanos focuses their hashtag use on advancing a neoliberal platform for gender equality-#FeminismoLiberal. Vox's hashtags lend more vivid context to the negativity observed in our earlier analysis of the party's gender-related tweets. The use of the tag #NoHablesEnMiNombre (#Don'tSpeakInMyName), is a particularly telling tactic wherein Vox attempts to shift critique towards progressive parties and politicians and feminist activists for "speaking in the name of" (hablar en nombre de) all women. Notably, this hashtag borrows some of the features of popular feminist twitter campaigns through a testimonial-style format in which Vox supporters share their experiences and state their opposition to feminism. The notion of a feminismo supremacista that crowds out the real voices of Spanish women and asserts itself within politics and public life against the will of the people is consistent with Vox's perception of an ideología de género that is corrupting Spanish institutions.
By parsing and visualizing the most frequently used hashtags by party, this study finds that Spanish parties use hashtags to frame their policies and critiques of other parties. By understanding these discursive frames, one can understand how parties align themselves with existing feminist dialogues on Twitter like #NiUnPasoAtrás, create contextual frames for their campaign promises as with #LaEspañaQueQuieresEsFeminista, and deflect criticism by attempting to change or destabilize the terms of feminist discourse through speech frames like #NoHablamosHacemos and #NoHablesEnMiNombre.

Conclusion
Each quantitative analysis undertaken in this study provides significant insight into the relationships between a party's interest in and position on key political issues and its Twitter rhetoric. However, these methods are most effective when put in conversation with one another and accompanied by close readings of tweets that contextualize noun and hashtag recurrence. By using word frequency analysis to assess the amount of attention a party dedicates to key electoral issues as a starting point and then parsing tweets further to identify co-occurrence networks, analyze sentiment, and conduct close readings, this study demonstrates the potential of quantitative approaches for focused analyses of online speech. In tandem, this set of tools reveals the connections between party stance, rhetorical tendencies, and affect.
Although the initial word frequency analysis found that gender was a topic of interest for all parties, it was the co-occurrence and sentiment analyses that unearthed a more detailed portrait of party variations in discussions of gender. The qualifiers and hashtags analyses demonstrated that, while right-leaning parties formulated their treatment of gender in ways that avoided overt affiliation with feminism, their tactics differed considerably, as indicated by the differences between Ciudadanos's desire for a moderate feminismo liberal, Vox's polemic threats of a feminismo supremacista, and the PP's near-total avoidance of the terms feminismo/feminista in favour of cautious statements about women's economic activity. Left-leaning Podemos and PSOE seized on this rhetorical indecision as a point of critique and an opportunity to assert their support for an unapologetic feminism sin apellidos (without modifiers) and comprehensive platforms for gender equality. This study's exploration of this conflict illustrates how distant methods-in this case, the use of co-occurrence networks to reveal overlap between the PSOE and Ciudadanos's use of the qualifier liberal for feminismo-uncover contested and inter-party conflicts in voter-focused communication. Furthermore, the gender case study reveals that all parties use disputes over the meaning of key terms like feminismo as opportunities to assert themselves against their political opponents.
While the PSOE sought to reinvent itself as a future-oriented feminist party with a favourable record on women's rights, the PP decried the PSOE's "politicization" of participation in the 8M feminist strikes and pointed, instead, to their own statistics for women's workforce participation as evidence of a commitment to gender equality.
Conversely, Podemos aligned itself with the feminist movement by supporting intersectional economic justice for women and dramatic policy changes to highlight its comparative progressivism in relation to the muted feminismo liberal of Ciudadanos and the avowed anti-feminism and gender ideology conspiracy theories of far-right party Vox.
Returning to this study's original question, then, is there a quantifiable relationship between the style and rhetoric of a party's Twitter speech, political platform, and political ideology? The findings of this analysis suggest that, yes, one can quantify the relationship between these elements, but it is most accurately measured when the quantitative analysis focuses on specific and often divisive topics such as gender equality, and with a multimodal approach that places analyses of word and hashtag frequency, co-occurrence networks, and sentiment in conversation with each other to reveal patterns in the rhetorical treatment of these topics. Given these findings, this study provides a model for similar studies of political Twitter in other geopolitical contexts and of other significant campaign issues, including immigration, regional autonomy, tax reform, and climate change.