A pragmatic view of hashtags: the case of impoliteness and offensive verbal behavior in the Brazilian Twitter

Studies on impoliteness have encompassed a wide range of social relations and scenarios, including some in which the means for achieving impoliteness is solely, or at least, more saliently linguistic. Among these scenarios, Digital Communication (DC) media, such as Twitter and blogs may incite linguistic manifestations of impoliteness. Since hashtags are ubiquitous in the interpersonal spaces of Twitter, the aim of this study is to investigate how hashtags were used to express offensive behavior and to convey impoliteness in the Brazilian Twitter. We compiled a corpus of 512 hashtags extracted from Brazilian Twitter in 2017 and 2018. Hashtags were manually collected from Twitter ‘trending topics’, and from 3 media sources, written in Brazilian Portuguese. Hashtags associated with offense, including derogatory language, taboo words and mockery were analyzed. Our findings suggest that hashtags served as strategies of impoliteness, since they intensified the contents of the tweets and framed the interpretation of verbal aggressiveness. While they did that, they also helped maximize face attacks in that they inscribed the tweets in a bounded communicative sphere of aggressiveness, mockery and derogatory language. Verbal attacks were mainly intended to politicians and to Brazilian public figures in general. While this permissiveness allowed for greater spontaneity and agility in the exchanges, it also encouraged outrageous uses, which would not be otherwise accomplished were it not for the transient and ephemeral framework of DC. This study shed light on the Pragmatics of DC, showing how ‘new’ features of the medium can be used to convey impoliteness.


Introduction
Aggressive behavior has been a major concern in social media. The digital environment is prone to disseminate offense, while it may also promote polarization and dissent (Oliveira & Carneiro, 2018, 2020Ott, 2017). Particularly from the viewpoint of offense, Haugh (2013, p. 47) suggests that " […] it can be understood as a social action initiated by the recipient in which he interprets the actions, or the conduct of the interlocutor (or of some other person or groups of people), as offensive". As one can see, for Haugh (2013), offense is a pragmatic act that is designed to (and also limited by) the type of social activity in which it arises. Taking this view into consideration, in this paper we aim to discuss how the hashtags employed in Brazilian Twitter operate to convey aggressive language, and thus act as strategies of impoliteness (Culpeper, 2005;Culpeper & Hardaker, 2017;Kádár & Haugh, 2013). In order to do that, we first present a discussion on the use of hashtags, followed by a presentation of the main theoretical framework that guides this study. After we do that, we report the methodology of data collection, and a sample analysis. These two sections are followed by a discussion of the most relevant data in the study, as well as by the final remarks.
Research on impoliteness has incorporated innumerous scenarios, particularly those in which the means for achieving impoliteness is predominantly linguistic (Kádár & Haugh, 2013;Culpeper & Hardaker, 2017). Digital Communication (DC) and Social Media in general, Twitter, blogs and Facebook are many times prone to displaying linguistic manifestations of impoliteness. Some of these manifestations fall under the label 'flaming'. Originally coined in the 80´ within users of social media, flaming is currently defined as any kind of " […] improper, provocative or aggressive contribution to any kind of forum, chat of any other kind of electronic communication platform targeting at unknown third parties" (Helfrich, Bedijs, Held, Maaß, & Uni Hildesheim, 2014, p. 298). In this study, however, we focus on impoliteness and aggressive behavior mostly targeted at public figures; for this reason, we do not consider that the term applies. From an interactional viewpoint, DC in general and social media in particular display a myriad of properties, all related to the sharing of social, cultural and communicative uses. These properties can be quite different from those identified in face-to-face communication (Marcuschi, 2012;Barton & Lee, 2013). As far as these differences are concerned, Marcuschi (2012) cites the lack of simultaneous feedback in some types of social media, the use of emoticons (and emojis) and the possibility of having multiple conversations when one is using the Internet, to mention a few of these properties.
Within the scope of DC, written genres emerge in combination with the features of oral production and with non-verbal material. This hybrid feature is widely accepted in the literature of the area and it motivated a full reconceptualization of the so-called opposition between oral production and written communication. As a consequence, a lot of non-verbal material is currently considered in the analysis of interactions taken place in DC (Barton & Lee, 2013;Halliday & Matthiessen, 2014). DC also makes room for a more intimate and closer-to-speech kind of language, which can very often embrace vagueness (Barton & Lee, 2013). Despite these inherent characteristics, Halliday and Matthiessen (2014) also call attention to the fact that virtual communication does not neutralize the differences between speech and writing. On the contrary, DC gives rise to the creation of certain conditions for the greater 'integration' of verbal and nonverbal language, for example, in the use of emojis, stickers, pictures and hashtags (Crystal, 2011;Wikström, 2014;Scott, 2015;Evans, 2017). In the next section, we will revisit the research on hashtags, and on their use in Twitter, as to provide the reader with a broader view of the topic.

On the use of hashtags on twitter
Twitter is considered one of the most popular online social media particularly because it allows for users to swiftly share information and opinions in real time. According to Han, Chung, Kim, Lee and Kang (2014), Twitter users post messages that are a response to the question 'What is happening now?' This response, initially limited to 140 characters, can currently reach 280 characters. The short messages in Twitter, the tweets, reflect the relevance of real-world events from the users' point of view. Tweets then appear on the feed of those who have decided to follow a given Twitter account. They may also be found through Twitter's search interface.
Hashtags are thus considered a main feature of Twitter. Users only need to use the #symbol, followed by a word and phrase, to form a hashtag. The primary function of hashtags is to tag information, making it easier to retrieve content (Scott, 2015).
Moreover, hashtags commonly help offer 'contextual clues' in DC by reinforcing the meanings to be communicated. Although hashtags do not work as conventional words, they can often replace words, phrases, or sentences, playing an important role from the semantic-pragmatic point of view (Scott, 2015). From this perspective, interactions mediated by the digital environment tend to somehow seek to guarantee the intensity and the authenticity of the communicative exchanges, compensating, in some way, for the lack of face-to-face contact (Wikström, 2014;Scott, 2015;Recuero, Zago, Bastos & Araújo, 2015;Oliveira & Carneiro, 2018). Hence, hashtagging is seen as a byproduct of the reduced affordances of the characterconstrained mode of the Twitter that fosters "[…] new kinds of meaning-making in language" (Zappavigna, 2015, p. 16).
As a consequence, hashtags are mainly employed in Twitter to organize, to share and to highlight discussion topics. They also help the reader to form implicated conclusions from content of the posts (Scott, 2015), allowing for certain meanings to be expressed and/or highlighted. Hashtags also mark an opinionated nature of communication, which can be at one time spontaneous and informal, as well as ironic or/ and insubstantial (Oliveira & Carneiro, 2018). Hashtags may also be associated with face-work and impoliteness, particularly when they contribute to maximizing potential attacks on the positive image of the other. (Oliveira & Carneiro, 2018) With regard to digital communication, Ott (2017, p. 60) contends that twitter discourse style is "[…] simple, impetuous and frequently denigrating and dehumanizing […]" when compared to discourse on TV, considered by the author to be "[…] silly, ridiculous and impotent" (Ott, 2017, p. 60). But not all Twitter discourse is harmful. Most of it is 'relatively innocuous'. It reflects the user´s´ narcissistic profile when posting messages that communicate very little. For this reason, it can be considered vacuous and insignificant. Such kind of discourse comprises about 80% of all tweets, while the remaining 20% can be considered harmful and/or offensive because "[…] issues of social, cultural, and political import are filtered through the lens of twitter, for twitter infects public discourse like a social cancer" (Ott, 2017, p. 60). According to the author, the kind of language used in harmful tweets is uncivilized, and it can promote fanaticism. It can also contribute to human insensitivity and disdain, especially with regard to public figures and to specific social groups or minorities.
Twitter fosters incivility, that is, communication that is impolite, insulting and/or offensive, as a result of its own features. According to Ott (2017), Twitter's informality is one such factor. The use of formal language may lessen the chance of offensive or insulting communication. The author contends that because Twitter is informal, "[…] lack of concern with proper grammar and style undermines norms that tend to enforce civility" (Ott, 2017, p. 62). In addition to this, since Twitter's users write to an imagined audience, the interaction that follows it tends to be depersonalized, with no concern as to how it will affect potential readers (Yus, 2011;Ott, 2017).
The tagging behaviour found in Twitter allows tweeters to create and to share multiple resources 'through tags'. Since these resources can be multiple, their "[…] meaning is socially constructed […]" (Recuero et al., 2015, p. 2), as it happens with social interaction in general. It thus can be said that the functions of hashtags may go beyond tagging information. In fact, several studies have investigated the affordances and uses of hashtags in tweets. A study conducted by Wikström (2014) categorized hashtags into eight communicative uses: (1) topic tags, (2) hashtag games, (3) meta-comments, (4) parenthetical explanations / additions, (5) emotive use/ emphatic use, (7) humorous use, and (8) memes and references to popular culture. The author also postulates that hashtags are used as a substitute "[…] for features that Twitter lacks, e.g., tagging instead of bolding or italicizing (Wikström, 2014, p. 148)". From these types, the ones that were more salient in the study were meme hashtags. In addition, hashtags can also be used in unconventional ways, such as in the punch line of jokes, or to create Internet memes that usually refer to elements of the so called 'pop culture'. In the present study, we confirmed the applicability of some of the categories proposed by Wikström (2014), particularly the ones containing meme hashtags, as we will attempt to outline further in this text.
Memes refer to the phenomenon of rapid spread of sentences, images or videos. Wikström (2014, p. 146-147) explains that meme "[…] has come to be used online to refer to phenomena such as particular phrases or genres of images or videos that start replicating rapidly throughout an online population". The author highlights that memes are less transient than hashtag games, and may refer to responses that are used over time. As an example, the phrase 'cool story, bro,' is used "as a sarcastic response to a story that is deemed pointless or boring." The phrases that are part of meme hashtags may have circulated online in a variety of contexts, and may have been seen in films, such as A few good men, in the following example (Wikström, 2014, p. 148): (1) Did I mention that our pardon system has a 96% success rate, i.e. former inmates rehabilitated and NOT reoffending.#youcanthandlethetruth .
In the Brazilian context, hashtags used as memes included #Fora (#Out or #No), followed by the name of a public figure, especially from Politics (Oliveira & Carneiro, 2018). Such phrase is commonly seen in street demonstrations throughout the recent history of Brazil (Figure 1), and has circulated in a variety of contexts and media. Another example is caguei (I do not give a shit), which is said to express disdain to a fact or person. Other memes included the name of a politician followed by 'injail' (nacadeia), as in #LulaeDilmaNaCadeia (#LulaandDilmaInJail) (Oliveira & Carneiro, 2018). As for the meta-comment hashtags, Wikström (2014) highlights that they can be used as hedges that suggest face-work. As for the use of hashtags for parenthetical explanations, the author points out that they help provide background information in order to clarify/disambiguate the illocutionary force of the utterances. Alongside these uses, emotive tags are typically employed to strengthen or change the illocutionary force of an utterance, similarly to what is seen in the use of non-verbal clues in face-to-face interaction.
Other studies have also investigated how hashtags contribute "[…] as a guide to the reader […]", triggering inferential processes for utterance meaning interpretation (Scott, 2015, p. 13). Scott (2015) points out that hashtags used to achieve this goal are generally placed after the main content of the tweet, and thus are not usually fully integrated into the main sentence. An example of such a usage can be seen (2): (2) @user I had a run in today... Like a pack of wolves when they all join in.#franklymydearidontgiveadamn (Wikström 2014, p. 148) In sum, from a pragmatic viewpoint, previous studies have shown that hashtags can be used to add context to the utterances, to lead the reader towards the formation of implicated conclusions and to help Acta Scientiarum. Language and Culture, v. 42, e50500, 2020 them identify salient referents (Scott, 2015). Hashtags may also convey emotions, by allowing for the expression of an opinion on a given topic in Twitter (Wikström, 2014). Furthermore, hashtags can also operate as a call for action, as in the examples from the political demonstrations in Brazil in 2013 (Recuero et al., 2015). Apart from these uses, our hypothesis in this paper is that hashtags can also be employed as strategies of impoliteness within a derogatory language use. In the next sections, we will discuss the issue of impoliteness and of language to cause offense in more detail.
(Im)politeness and face work From Goffman's (1976) perspective, face work refers to the linguistic and non-linguistic actions performed by participants to claim their social values or to maintain their self-image in a way that is considered satisfactory for the interaction on focus (Haugh, 2013). According to Goffman (1995), there are two main processes, related to face-work: the avoidance process, which consists in avoiding social threats, and the corrective process, which acts as to repair possible damages to the face, or to the public image of the speaker.
Complementary to the notion of face, the concept of territory refers both to the physical territory, or to the "[…] portion of space that surrounds an individual […]" (Goffman, 1973, p. 44), and to body parts, clothing and personal objects. In addition, the notion of territory encompasses the reserved domains of the conversation, that is, the right of the individual to control who can address the speech or the right of one group of people to protect themselves from the intrusion and indiscretion of others (Goffman, 1973). Brown and Levinson (1987) revisit the concept of face-work in Goffman mostly stemming from the assumption that the politeness phenomena can be understood on the basis of a 'Model Person': "[…] a willful fluent speaker of a natural language', endowed with rationality and with two face wants, that is, a 'negative face' ('the want to be unimpeded) and 'positive face' ('the want to be approved of in certain respects')" (Brown & Levinson, 1987, p. 58, grifo nosso). A major tenet of Brown and Levinson's (1987) model is the Face Threatening Act (FTA), in that they are associated with "[…] certain kinds of acts intrinsically threatening the face" (Brown & Levinson, 1987, p. 65). FTAs are classified according to the kind of face threatened (positive or negative) and to whether the threat concerns the hearer's face or the speaker's (Brown & Levinson, 1987, p. 65-68). For example, requests threaten the negative face of the hearer; criticisms attack the positive face of the hearer; thanks threaten the negative face of the speaker (as a debt is acknowledged). FTAs are thus a central notion in Politeness Theory. In this respect, Leech (1983) conceives of them as "[…] a constraint on human behavior that makes us, on the one hand, avoid discordance or communicative offense and, on the other, maintain or increase communicative courtesy or courtesy" (Leech, 1983, p. 64).
Along the same lines, Oliveira, Cunha, and Miranda (2017) claim that the notion of face-work under Brown and Levinson (1987) underscores an important conceptual slip. On the one hand, it is restricted, since it only corresponds to the use of linguistic devices that mitigate the threat of speech acts. On the other hand, the notion is broadened, since it goes on to encompass the strategies used to lessen the attacks on the negative face, and not only on the positive one.
As for the Politeness theory, Kienpointner and Stopfner (2017) bring to the fore the fact that the description of face and wants in Brown and Levinson (1987) tends to overestimate the importance of conscious choice and of rational, goal-oriented practical reasoning for verbal behavior. The authors point out that, while Politeness may sometimes be seen as biased towards an obsession with individual rights and wants, Impoliteness cannot be simply taken as a 'destructive deviation' from Politeness (Kienpointner & Stopfner, 2017). Kienpointner and Stopfner (2017) also call attention to the fact that impoliteness may not always be the marked, exceptional counterpart of politeness. In fact, impoliteness can be frequent, and, in some institutional contexts, it can even be the norm.
Furthermore, the tenor of the notion of impoliteness is social interaction and immediate context. For Culpeper (2005), impoliteness is put to rise when the speaker communicates face attack intentionally, when the hearer perceives and/or constructs behavior as intentionally face-attacking, or when there is a combination of two (Culpeper, 2005). Along these lines, Culpeper (2005) states that positive impoliteness is the use of strategies designed to damage the addressee's positive face wants, for example in the use of unconcerned or unsympathetic expressions, inappropriate identity markers, obscure or secretive language, taboo words and name-calling.
Negative impoliteness, on the other hand, emerges from the use of strategies designed to damage the addressee's negative face wants, for example, to frighten, to scorn or to ridicule, to be disdainful, to belittle the other and to invade the other's space (literally or metaphorically) by explicitly associating the other with a negative aspect or by putting the other's indebtedness on record. According to this view, sarcasm and scornful language also help produce impoliteness, since they are typically taken as offensive (Kádár & Haugh, 2013;Dynel, 2015).
Impoliteness is also a cognitive phenomenon. This means that no linguistic forms can be impolite per se: "[…] no linguistic form or strategy can be inherently more or less polite" (Kádár & Haugh, 2013, p. 143). This is partly the reason why, in more recent research, the term 'culture' has been replaced by the concepts of 'community of practice', as a starting point for impoliteness analysis. According to Mills (2003), the term refers to "[…] a loosely defined group of people who are mutually engaged on a particular task" (Mills, 2003, p. 30). The use of aggressive language and insult was also studied by Jamet and Jobert (2013), when conducting research in the digital environment. Regarding this, the authors discuss a kind of fascination Internet users have with taboo words. From this perspective, offensive language can be seen as a form of transgression of valid rules, and as a way of constantly saying 'no' (Jamet & Jobert, 2013). According to these authors, the use of taboo words can mark dissent, as well as it is typically used to express aggression and to highlight conflict in online communication.
Having traced this theoretical panorama of the research, we set off to describe the procedures used for data collection and for data analysis.

Data collection
In order to investigate how hashtags were used to express offensive behavior and to convey impoliteness, 530 tweets were collected in Brazilian Twitter. Data was gathered from September 2017 to November 2018. Posts containing hashtags were collected from t he 'trending topics' and from 3 media outlets, all written in Brazilian Portuguese. These included a magazine (Vejahttps://veja.abril.com.br/), a newspaper (Estado de Minas -https://www.em.com.br/) and a TV/newspaper (G1 -https://g1.globo.com/) Twitter accounts.
Taken that random sampling refers to a variety of selection techniques in which sample elements are collected by chance for analysis (Emerson, 2015), this procedure was considered a critical component in the initial phase of our study in order to guarantee that the data gathering was unbiased. The criterion for inclusion was the presence of a hashtag within the body of a tweet. We searched for hashtag occurrences among the 'trending topics' and in posts published by the three media sources. According to Twitter help page (2020), a trending topic is a term coined by Twitter to refer to the most used keywords on the social network during a given period of time. Once posts containing hashtags were found, they were collected for further qualitative analysis. The communicative context and/or the trigger for posts were also stored, so as to preserve the interaction flow.
As a further step, a general quantitative analysis of the corpus was also performed, so as to determine which hashtag types were frequently found. In order to do that, hashtags were manually extracted from the tweets and a simple and descriptive statistical analysis was performed. In order to be primarily categorized as impolite and/or aggressive, the hashtag needed to contain offensive and/or taboo words, such as 'fuck, asshole, bitch, shithole', as discussed in Oliveira & Carneiro (2018) data analysis. In addition, hashtags containing 'no, out or neverAgain' were also included, since they were also used to cause offense.
The corpus was then qualitatively analyzed. The hashtags found in tweets were described in terms of their offensive potential, following the framework proposed in Culpeper (2005), Dynel (2015) and Culpeper & Hardaker (2017). In the sample analysis selected for this paper, tweeters had their photos wiped off from the posts, which was done with the purpose of preserving their identities.

General results overview
A total of 512 hashtags were grouped into those containing insults, taboo words, 'no' and 'out'. Some of these hashtags also fell into the category of meme hashtags, as it was also identified by Wikström (2014). A primary quantitative analysis of the corpus revealed that hashtags containing out and the name of a public figure totalized 20.11% of the data, while hashtags containing no and the name of a public figure amounted to 12.5%. These figures accounted for almost a third of the data analyzed here. As for insults and offensive language in general, they accounted for 50.5% of all data collected, of which taboo words represented 16.7% of offensive posts. These findings are illustrated in Figure 2: Following this initial quantitative analysis (Figure 2), the qualitative analysis was carried out. The content of the posts and the hashtags were manually analyzed, as we will see in the examples that follow. These examples were randomly selected from each type of hashtag found in our data.

Data analysis: meme hashtags, taboo words and offensive language use
About a third of our data was associated with the creation of memes, which refer to the phenomenon of the rapid spread of sentences, images or videos over Twitter. Meme hashtags usually exhibit a transient nature, since they do not always trigger responses from tweeters. In our data, meme hashtags typically contained derogatory language, for example, #Fora (#Out or #No), followed by the name of a public figure, especially from politics or people from the judicial system. Figure 3 illustrates this use, in which the hashtag was employed to invoke the resignation of a Supreme Court judge. The post translates as: '#OutGilmarMendes no use complaining, need to take him out of the Supreme Court'.  4 contains #fora (out), used here to call out for the resignation of three public figures, which were all in office at the time of data collection. In our findings, this hashtag was typically used to manifest a strong feeling that some public figures be kept away from the political scenario. The hashtag #fora (out) was targeted at politicians from various political parties, and not at a single party in particular, as one could have initially expected. The post in Figure 4 roughly translates as, 'Did Adalclever, the son of Federal Legislator Mauro Lopes, become poor with politics? Shameless man, he was an ally to Governor Pimentel as he led the State of Minas Gerais to bankruptcy. #OutAdalclever, #OutPimentel, #OutDilma'. As Figures 3 and 4 demonstrate, meme hashtags about politicians were very prominent in our data. These hashtags licensed disparaging commentaries about the political class as a whole, as they instilled a belief or an opinion that was detrimental to their reputation. As they did that, hashtags acted as strategies of impoliteness, marking offense as the norm. This type of derogatory language use was also identified in hashtags such as '#injail', usually followed by the name of a Brazilian politician, as illustrated in Figure 5. As one can see, the hashtag conveyed on record impoliteness and offence, once the tweeter overtly communicated the face attack intentionally (Culpeper, 2005), while he/she also expressed his/her wish that a former president and three other politicians of national projection were taken into custody. Another interesting point about hashtag use was the fact that, in approximately 66% of our data, taboo or insulting language (Culpeper, 2005) was observed. Particularly with respect to taboo words, they accounted for 16% of the data analyzed. Along these lines, #Caguei (#donotgiveashit) was considered prominent in our corpus. It manifested disdain and indifference towards pieces of information being communicated by national news outlets. As one can see in Figure 6, the hashtag acted as a strategy of impoliteness, in that it damaged the addressee's positive face wants (Goffman, 1976;Brown & Levinson, 1987;Culpeper & Hardaker, 2017), while serving as a potential attack to the reliability of the mainstream media. The full post translates as 'Man, I almost cried at the news. What if I had seen the scene myself? #donotgiveashit'. Figure 7 contains a tweet in which the user reacted to the news that the front runner in 2018 Brazilian presidential campaign (who later became the President) was recovering from a stab. In order to manifest disdain, the tweeter used the hashtag '#FodaSe!' (#FuckYou!), followed by Ninguém quer saber (nobody cares!). As one can see, the use of taboo words was frequent and mostly targeted at national news outlets or at public figures.  Figure 8 is another example of taboo words used to attack the face of those that oppose the fact that a former president was arrested. In reaction to the news that supporters of the former president were organizing a New Year's Eve celebration close to the area where he was kept in captivity, the tweeter wrote: Vai ter lugar para todo mundo? E Pode? The post translates as: 'Will there be enough room in his jail for everybody? Is this permitted?' This content is followed by the hashtag #LulaTaPresoBacaca and PTNuncaMais, which translate as '#LulaIsArrestedAsshole' and '#PTNeverAgain'. The tweeter is not only addressing those that opposed the imprisonment, but also he/she is also attacking the former president´s political party as a whole (PT, which is short for Partido dos Trabalhadores -Worker´s Party). Acta Scientiarum. Language and Culture, v. 42, e50500, 2020 Figure 9 also illustrates the use of hashtags with aggressive language targeting at a public figure. The tweet came as reaction to the declaration of a member of the Brazilian Supreme Court, Judge Toffoli. The judge stated that 'people should care about the sin, not the sinner', as a way to display his support to the issue of prisoner rights. As a reaction to this declaration, the tweeter wrote: Falou o ministro que anda com carro blindado e com vários seguranças #calabocatoffoli (look who's talking: the judge that rides an armored car packed with security guards #ShutUpToffoli). In addition to the previous part, which is by itself impolite and disdainful, the hashtag '#ShutUpToffoli' is impolite and offensive, as it directly attacks the personal space of the Judge, and morally attacks his public image (Dynel, 2015). By doing so, although the tweeter was not probably expecting the judge to read his post, he steeply manifested his discontentment with what had been declared by the judge in the news. In this example, the hashtag signaled an aggressive way to say 'no' and to express dissent, as Jamet and Jobert (2013) pointed out in their work on aggressive linguistic behavior. The examples shown so far evidenced the impolite and opinionated nature of Brazilian tweets, intensified by the use of hashtags (Page, 2012;Oliveira & Carneiro, 2018). They also indicated that Brazilian Internet users made use of hashtags not only to classify or to organize the posts in their respective topics, but also to express meanings that went beyond the limits of the actual posts. From this perspective, we also observed that the tweets analyzed in our corpus contained hashtags that were frequently used in combination with images, emojis, and other icons. These features were employed in order to reinforce intended meanings. This is also in accordance with the multimodal nature of DC (Crystal, 2011).
In the next section, we will discussion our finding in more details.

Discussion
The data analyzed here confirmed the applicability of some of the categories proposed by Wikström (2014), particularly concerning memes hashtags. In our corpus, these hashtags, which included expressions like 'no' and of 'out', followed by a proper name, were mostly used as strategies to ridicule political parties, members of the judiciary and politicians, accounting for almost 30% of the data. While they did that, hashtags also expressed the tweeters stance regarding the behavior or ideas of such targets. Moreover, they were also employed to convey mockery and dissent.
Most of the data analyzed here comprised insults and the use of taboo words, with taboo accounting for 16.8% of the corpus. Insults included words and expressions such as 'corrupt, coward, thief, assassin, fool, rubbish, loser', to name a few. Similar to meme hashtags, the use of taboo words and insults was targeted to public figures, particularly to politicians. In addition, these hashtags were also employed to attack pieces of national news and the mainstream media, for example, with the use of expressions like 'fuck you, do not give a shit' and 'asshole'.
In general, our findings indicated that hashtags acted as strategies of impoliteness, allowing for meanings to be communicated in accordance with: (a) the opinionated nature of tweets, commonly linked to a broader topic previously initiated on Twitter; and (b) the spontaneous, informal and offensive nature of these commentaries, which were often superficial and insubstantial.
As a consequence of these characteristics, the use of hashtags in our corpus conveyed impoliteness as the norm. This finding was evidenced, for example, #Tomavergonhanacara (#shameonyou) and "#caguei (#doNotGiveaShit), which were prominent in our data. In fact, around 66% of the hashtags we collected contained taboo words and insults, including expressions such as #foda-se (#fuck you), #ladrão (#thief), #corrupto (#corrupt), as well as commands, for example, #fora (#out), #calaaboca (#shutup) and #na cadeia (# injail). These hashtags were mainly addressed to politicians and to public figures in general. In addition, commands demanding that public figures left their positions accounted for almost a third of our data.
Particularly with regard to insult and to language used to cause offense, the belief in anonymity seemed crucial in our data, since Internet users acted as if they could never be held accountable for the offenses they published. In addition, offensive language was also prompted by the lack of a face-to-face contact and by the informality of the medium, as suggested by Ott (2017). At the same time, aggressive language also led to a very superficial kind of argumentation, in which reasoning was many times overshadowed by excessive insult, and/or by mere aggression. In this sense, it reflected the transient and swift nature of the cybernetic environment, while providing the reader with the contextual clues needed for communication to be considered effective.
More importantly, our findings challenged the concepts of facework (Goffman, 1976) and of politeness (Brown & Levinson, 1987) to reaffirm the notion of linguistic impoliteness, as claimed by Culpeper (2005), and by Culpeper & Hardaker (2017). Attacks on the positive face of public figures were achieved mostly by means of meme hashtags, containing the expressions out and no used in the sense that such figures were being regarded by the tweeter as 'personae non gratae'. By the same token, attacks on the negative face of the targets were achieved by name calling or insults, thus putting the target´s indebtedness on record.

Final considerations
In this study, it was possible to identify that hashtags framed the interpretation of the opinions in the posts analyzed. They also helped perform the attacks on the negative and on the positive faces of the other, in that they inscribed the tweets in a bounded communicative sphere of aggressiveness and insult. From this perspective, hashtags instilled offensive and impolite verbal behaviors, notably pervasive in the virtual environment. While this permissiveness allows for greater spontaneity and agility in the exchanges, it also encouraged outrageous uses, which we claim would not be otherwise achieved were it not for the transient and ephemeral framework of DC. Furthermore, the Brazilian tweets analyzed here allowed for this digital text to be described as opinionative, ephemeral and widely circumscribed to a specific communication situation, in which impoliteness was the norm. Furthermore, within the use of hashtags, politeness was many times replaced by unjustified offense and by verbal insult.
More specifically, hashtags on Brazilian Twitter were employed to convey a form of derogatory and aggressive language, mainly intended to attack Brazilian politicians, pubic figures and the mainstream media. From this viewpoint, our findings also reaffirmed the results of Jamet and Jobert (2013), concerning the fascination with taboo words, and with the use of derogatory language to express dissent. With respect to this, Ott (2017) has shown that uncivilized language was found in Twitter in 20% of all tweets that he analyzed. Our data has revealed that taboo words accounted for 16.8% of aggressive language, while half of it contained insults of other types. It should be remembered that our corpus contained only hashtags with offensive language and cannot be fully compared to Ott's data. Additional studies with a larger corpus should be carried out to confirm this trend, which we believe is an important issue to be further investigated.
Finally, we believe that this topic was not exhausted by the considerations that we made in this paper, and therefore can be expanded to other social media and to different populations, as an attempt to reinforce or to relativize the results found here. Although this study may have limitations, for example, it solely focused on one kind of social media (Twitter), we hope we may have taken a step forward in a field of research that is nearly unexplored, though very intriguing and vast.