OUR SAVIOUR WILL NOT BE A WOMAN: USERS’ POLITICAL COMMENTS ON YOUTUBE TALK SHOW OFFICIAL CHANNEL

ABSTRACT

This article aims at studying the representation of four candidates (two male and two female) in the last Brazilian Presidential elections in a corpus of comments posted online. The comments were posted as part of the reception of a series of interviews aired and broadcasted through official YouTube Channel by a top-rated talk show in a Brazilian Public Television. The theoretical framework lays on Systemic-Functional Grammar. Online Data scraping software collected data anonymously. All comments were processed using software for corpora management, and quantitative results were plotted using Gephi. The results show that male candidates tend to be judged by their political options and social-political origin. They represent political saviours or possible political menaces. Likewise, women are judged for their political and ideological affiliations. However, their physical appearance is continuously evaluated as they are subjected to sexist and sexualised comments.

systemic-functional linguistics; women; appraisal; transitivity; lexis

RESUMO

Este artigo tem como objetivo estudar a representação de quatro candidatos (dois homens e duas mulheres) nas últimas eleições presidenciais brasileiras em um corpus de comentários publicados online. Os comentários foram postados como parte da recepção de uma série de entrevistas veiculadas e transmitidas pelo canal oficial do YouTube por um programa de entrevistas entre aqueles com a melhor classificação em uma televisão pública brasileira. O referencial teórico é a gramática sistêmico-funcional. O software de raspagem de dados on-line coletou dados anonimamente. Todos os comentários foram processados usando o software para gerenciamento de corpora, e os resultados quantitativos foram plotados utilizando aplicativos específicos. Os resultados mostram que todos candidatos tendem a ser julgados por suas opções políticas (direita ou esquerda) e origem (região). Todavia as mulheres também são julgadas por sua aparência física, sendo submetidas a comentários sexistas e sexualizados.

linguística sistêmico-funcional; mulheres; avaliação; transitividade; léxico

Introduction

This paper analyses the comments of YouTube users to the interviews of four 2018 Brazilian presidential pre-candidates. The interviews were promoted and broadcasted by Roda Viva , a prestigious talk show in Brazil. The data analysis will focus on comparing comments on the two most commented male pre-candidates (Jair Bolsonaro and Ciro Gomes) to the two only female interviewees (Manuela D’Ávila and Marina Silva).¹ The main objective is to analyse whether a gender-oriented culture is going to influence the candidates’ representations ( THOMPSON; THETELA, 1995 ; DELU, 1991 ). Systemic-Functional Linguistics, henceforth SFL ( HALLIDAY, 1978 ; HALLIDAY; MATTHIESSEN, 2014 ) is the main theoretical background and Corpus Linguistics ( BAKER, 2009 ; BAKER; MCENERY, 2015 ) and Network Science ( BARABÁSI, 2002 ; SCOTT, 2013 ) are the core methodological approaches.

Roda Viva is a traditional Brazilian talk show, which has been broadcasted by TV Cultura — a public TV supported by the State of São Paulo — since the mid-1980s. Roda Viva features a host mediator and some journalists from different Brazilian mass and independent media. The guest is placed at the centre of a circle and surrounded by the journalists, who are in a higher position, so the guest has to turn his/her chair around to answer the questions. The general model is different from a typical interview and closer to an everyday conversation since interruptions are frequent. The host organises the many interventions and makes sure time is equally distributed.

The talk show has reached elevated consideration amongst the Brazilian public and journalists, due to the high profile of their guests and quality of the interviews. Roda Viva has regularly interviewed the candidates for executive offices before the general and mid-term elections since Brazilian re-democratisation. In 2018, Roda Viva kept this tradition and invited some of the presidential candidates to the talk show during the first and second rounds of Brazilian elections. The talk show interviewed ten candidates in the first round (eight men and two women) and one (male) in the second round. After a live transmission, TV Cultura makes each interview available in Roda Viva’s YouTube channel. As with any YouTube video, users might take their time to comment and discuss the outcome of such interviews.

The smaller number of women at Roda Viva’s interviews might be a consequence of the lower levels of women participation in Brazilian politics. Coelho and Batista (2009) observe that such an issue seems to reflect the roles women should fill in a male-centred society. They point out that women have been regarded as “naturally inferior” and unable to fulfil public offices without compromising their primary function, which is to take good care of their family, since the first Brazilian Constitution in 1891. For the authors, the struggle for more political participation reflects an attempt to leave the privacy of a family home, and, consequently, the roles of mother and housewife.

Brazil is now the 161^stcountry in terms of female presence in public offices ( ROSSI, 2018 ). Even though women have been able to vote in Brazil since 1932, the first female elected to the House of Representatives being Carlota de Queirós in 1937, their participation seems to have not changed much since the 1940s. The first female governor was Roseana Sarney, elected in 1994 and the first woman president was Dilma Rousseff ( GUERRA, 2016 ), impeached by a civil coup in 2016. There has been less than 4% of women in ministerial positions and secretaries ( ROSSI, 2018 ), their number in city councils is below 15%, with an average of 7 male counsellors for each lady ( SANTOS, 2017 ); 61% of the Brazilian mayors are white males. The average participation of women in public offices is below 20% in all Brazilian states, both in 2014 and 2018, only one female governor was elected ( FLORENTINO, 2018 ).

Some official regulations try to guarantee equal participation of men and women in Brazilian politics. Firstly, a 1996 law imposed a percentage of 20% of women on every list of candidates. A 1997 bill increased this percentage to 30% from the year 2000 on, with a minimum of 25% for the 1998 elections. However, parties do not give the same emphasis and importance to the female candidates, as the best part of them is in the party list only to fulfil the legal quota ( PINTO, 2001 ). Such a political context seems to contribute to the relevance of this study. It is in this context that it would be essential to understand how the ordinary YouTube public represents women’s political participation in Brazil. We are interested in observing whether women who are running for the most important political office in the country are treated differently from their fellow male candidates.

As other studies have pointed out ( LIMA-LOPES; PIMENTA, 2017 ; GABARDO; LIMA-LOPES, 2018 ; LIMA-LOPES; GABARDO, 2019 ), a conservative opinion towards the role of women in society might be expressed through online communities, especially social media. It is possible to observe that the anonymity is a critical component when expressing online sexism ( FOX; CRUZ; LEE, 2015 ), a fact that might include YouTube as a place for the expression of uncivil comments ( HALPERN; GIBBS, 2013 ; MOOR; HEUVELMAN; VERLEUR, 2010 ). This paper aims at contributing to the growing research on YouTube ( ANDROUTSOPOULOS; TEREICK, 2015 ), as it discusses discursive and political opinions on users’ comments.

The primary theoretical approach lies in Systemic-Functional Linguistics ( HALLIDAY, 1994 ). This perspective makes it possible to unveil the relationship between lexicogrammatical and meaning choices, as it brings to this study a framework that relates semantic and structural strata in a paradigmatic axis ( THOMPSON, 2004 ). Network Science ( BARABÁSI, 2002 ; SCOTT, 2013 ) and Corpus Linguistics ( STUBBS, 1996 ) played an essential role as they allowed the processing and visual representation of large amounts of linguistic data.

The study of feelings and emotions has become a growing topic in many corpus-driven studies ( TABOADA, 2016 ). The application of such methodologies on appraisal, however, has shown the need for the development of specific models ( TABOADA; GRIEVE, 2004 ). This is so because the appraisal model mixes binary choices — such as attitude or engagement, affection or judgement — and continuous choices — force and focus. On the other hand, the SFL model is somewhat dependent on the context, generating the need for identification of larger units, such as whole nouns or verbal phrases ( TABOADA; GRIEVE, 2004 ). Lima-Lopes (2020) developed a framework to analyse how UK conservative media characterise immigrants during the days that preceded Brexit in 2019. He agrees with Taboada and Grieve (2004) , as his results show that different newspapers tend to have their own evaluative profiles, as each newspaper shows different appraisal strategies, associating them in a unique way.

However, such studies have English as their main topic of study. A model for the automatic analysis of appraisal in Brazilian Portuguese is in its infancy. So, this paper intends to provide an initial qualitative-quantitative approach to map resources regarding possible gender-biased evaluation in order to provide possible lexical classification parameters for future studies.

The next section discusses the theoretical background of this research, and it is followed by the methodological procedures for data collection and analysis. Section 4 is centred in the results and discussion.

Theoretical Background

Systemic-Functional Linguistics ( HALLIDAY, 1978 ; HALLIDAY; MATTHIESSEN, 2014 ) is an approach known by its conception of language as a semiotic system ( MARTIN, 2016 ), in which the relationship between content, form and expression is represented as a hierarchy. Here the grammar is seen as a set of possible choices deeply influenced by the context, as the meaning is functionally motivated ( THOMPSON, 2004 ).

SFL is an approach which understands grammar not only as a body of possible structures but as a means of organising the appropriate language choices in any interaction. Such choices reverberate, according to Halliday (1978) , our social and individual experiences. When a speaker uses a wording, she/he is performing a social act. Such an act is influenced by both the Context of Culture and the Context of Situation. As a result, any language choice represents our relationship with others, our representation of the world, as well as the way we structure and organise our messages ( THOMPSON, 1996 ). The Hallidayan socio-semiotics couples meaning and structure as it explains how we make socially meaningful choices. The grammar is a paradigmatic system that allows the speaker to create combinatorics of signs which have a specific sense within a communication situation. The organisation of the structural constituents is a result of a demand for the context and the functions they perform ( HALLIDAY; HASAN, 1991 ; KRESS, 2010 ). In other words, as we choose to say something in a particular way, we are also “not choosing” some other possible wordings, which would bring about a different set of meanings ( THOMPSON, 1996 ).

Thus, the texts may vary concerning their production context and their social objective. The genre (or Context of Culture) would be responsible for the choices regarding the more general social processes ( VIAN JUNIOR; LIMA-LOPES, 2005 ). On the other hand, the Register would work as a means of instantiation of the many genres, materialising our global cultural institutions into the specificities of a given context of communication ( EGGINS; MARTIN, 1997 ; MARTIN; WHITE, 2005 ). In other words, the Register would be responsible for our language choices instantiated by the lexicogrammar during the actual moment of communication. As a result, the Register imprints the particular characteristics of each situation on our language conventions. In this study, it is expected that some patterns might emerge from the corpus, representing the candidates according to the social and political views of the YouTube users who posted comments on each interview.

Halliday (1978 , 1994 ) introduces a three-dimensional model of analysis ( IEDEMA, 2003 ), as language choices actualise into texts, they are instantiated in terms of three language metafunctions. They are not the only metafunctions we perform in our daily communication, but are essential for our social, organisational and representational capacities ( HALLIDAY, 1994 ; HALLIDAY; MATTHIESSEN, 2014 ). Our choices are also conditioned for such metafunctions as they give rise to the social and ideological semantic-functional nature of the context.

In this article, I focus on the instantiation of language patterns in the System of Transitivity ( HALLIDAY; MATTHIESSEN, 2004 , 2014 ) and its relationship with the system of Appraisal ( HUNSTON; THOMPSON, 2000 ; MARTIN; WHITE, 2005 ).

Halliday (1994) points out that our most strong representation of experience lies within the acknowledgement of our life events, such as being, becoming, happening, sensing, and many others. The system within the grammar that would make it possible for us to carry it out is the transitivity, which would be responsible for aiding the speaker/writer to attribute meaning to her/his experiences and create a representational world.

In this system, responsible for translating our representation into language choices, a proposition would have three essential components: 1) the process, 2) its participants, and 3) circumstantial elements. The verbal component always represents the process; the participants are usually represented by nouns, which either carry out or are affected by the process; and the circumstantial elements are responsible for encoding the background against which a process takes place. Transitivity is, then, accountable for transposing both our inner and outer experiences into an actual set of processes types, participants and circumstances ( HALLIDAY, 1994 , 2004 ; HALLIDAY; MATTHIESSEN, 2014 ).

Processes would be better depicted as a continuum, a non-sharp colour matrix in which meanings blend into each other. The semantic-functional characteristic of language makes it almost impossible not to come across examples which are “in-between” and skip any categorical definition ( HALLIDAY; MATTHIESSEN, 2014 ). It is essential to keep in mind that Halliday’s model was developed based on English and such semantic borders are even blurrier in Brazilian Portuguese (BP). As Lima-Lopes (2017a) brings up, it is impossible to discuss the classification of processes in Portuguese without drawing the use of collocates as collocations influence profoundly in the meaning and function in many ambiguous cases. As the results will show, behavioural processes tend to be quite a different configuration in BP.

There are six kinds of processes according to the SFL. Material processes are responsible for representing our experience in the outer and physical world and instantiate concrete and real doings/goings-on. Mental processes stand for meanings and experiences that go precisely in the opposite direction, experiences in our inner world, instantiated in our consciousness, such as actions of sensing, feeling, and thinking, and are represented as language choices. Relational processes are known to represent meanings related to identification, attribution and ownership. Behavioural processes lie in the frontier between Material and Mental processes and actualise actions which have both an inner component and an actual output. Verbal processes are between the mental and the relational and can be described as responsible for symbolic relationships that take place in our conscience but instantiate through verbal language. In time, existential processes are between material and relational processes. They are related to any phenomenon recognised as existent, a material manifestation of a symbolic element.

According to Halliday (2004) , each proposition must have at least one participant. Participants are close to the processes; they are responsible either for bringing about its occurrence or by being affected by it in some fashion. The semantic nature of processes makes each participant perform specific roles. So, participants should vary according to the kind of process one chooses within each proposition.

Table 1 summarises participants according to their functions, based on the LSF base theory ( HALLIDAY, 2004 ) and in a taxonomy introduced by Lima-Lopes (2008 , 2014 ) when studying direct sales letters and Greenpeace bulletins in BP. The participants in the active role would be those which we would identify as responsible for the process in interpersonal terms ( HALLIDAY, 1994 ), they would be the first answer to the question: Who is responsible for (…) ? Modified participants, on the other hand, are those to whom/which the quantum of change represented by a process directs towards, they would be the first answer to a question like Whom has been affected by (…) ? Beneficiaries are those participants who benefit directly or indirectly from the action represented by a process. According to Halliday (1994) , one should not assume that all beneficiaries shall carry some positive semantic prosody, despite the name of the category. In BP, the beneficiary would be performing the ‘indirect object’ ( objeto indireto ) in terms of the traditional BP normative grammar ( ROCHA LIMA, 2010 ). Range, finally, are those participants who/which instantiate the scope of a verb; they sometimes are responsible for our very meaning projection.

Table 1
– Participants and processes

Function Process	Transitivity function
	Material	Behavioural	Mental	Verbal	Relational	Existential
Agent/ assignee	Actor	Behaver	Senser	Sayer	Possessor * Attribute ♠ Token (value) +	Existent
Modified/ assigned	Goal			Target	Carrier ♠ Possessed * Type +
Beneficiary	Recipient Client			Receiver
Range	Scope	Behaviour	Phenomenon	Verbiage

• Caption: *Relational possessive, ♠ relational attributive, + relational intensive
Source:Halliday (2004)Lima-Lopes (20082014

Lima-Lopes (2008 , 2014 ) points out that relational processes carry an important transitive feature in BP, mostly because other processes might instantiate them besides ser and estar (both translating as to be in English); examples would be ficar, tratar and oferecer. The first might instantiate both material ( Fique aqui – stay here) and relational attributive ( Fiquei calmo – I was calm). The second would instantiate verbal ( Tratamos do problema – we discussed the issue), material ( Tratamos de você – we took care of you) or relational ( Trata-se de um suplemento – it is a supplement). The latter might instantiate verbal ( Nós oferecemos um café – We offered him a coffee) or relational possessive ( As garantias que nosso seguro oferece – the coverage our insurance has). When such processes instantiate relational meanings and keep their grammatical structure, ascribing attributes, values and possession to the participants. The idea that a process might have its meaning determined only in context is especially true in BP, since collocational and semantic patterns are, sometimes, a useful tool for understanding the meaning of a process ( LIMA-LOPES, 2017a ).

Table 2 summarises the circumstantial elements. Circumstantial elements tend to occur freely with any kind of process, and they would, therefore, map onto adjuncts. Yet, they do not have the potential of instantiating subjects or of being identified as responsible for a clause ( HALLIDAY; MATTHIESSEN, 2014 ). This is because their primary function would be establishing the background in which the interaction takes place.

Table 2
– Types of circumstantial element

Type		Wh-item	Type		Wh-item
Extent	Distance	How far?	Contingency	Condition	Why?
	Duration	How long?		Default
	Frequency	How many times?		Concession
Location	Place	Where?	Accompaniment	Comitative	Who with?
	Time	When?		Additive	Who else?
Manner	Means	How?	Role	Guise	What as?
	Quality	How?		Product	What into?
	Comparison	What like?	Matter		What about?
	Degree	How much?	Cause	Reason	Why?
Angle	Source			Purpose	What for?
	Viewpoint			Behalf	Who for?

Source:Halliday and Matthiessen (2014

This is especially true if one takes circumstances of place and time as examples. However, it is a bit more problematic when we have a closer look in circumstances like behalf or accompaniment. In some cases, circumstances might be a way of bending the structure of a language to allow new participants or processes to be introduced into the clause. A phenomenon that Halliday calls indirect participants ( HALLIDAY; MATTHIESSEN, 2014 ).

The second SFL system that contributes to this research is the appraisal. The appraisal is a system aimed to describe our stance towards other people, ideas and representations. In this perspective, the language would be responsible for mirroring our value systems, through socially motivated linguistic resources and related to their context of use ( THOMPSON; ALBA-JUEZ, 2014 ).

Three subsystems are responsible for expressing appraisal meanings. The system of attitude represents our feelings as they are constructed in the text. There are three subsystems here: (1) Affect, responsible for the registers of positive and negative emotions; (2) Judgement, focusing on attitudes towards behaviours, which we can admire, praise, criticise, or condemn; and (3) appreciation, related to our aesthetic positions.

Engagement is a set of resources available to the writer/speaker to place her/himself towards the values present in the text, as well as to the people he or she addresses. In this sense, a text would be the result of the interaction with several other texts and voices and the engagement provides a framework to systematically account for the way our positions are linguistically represented ( MARTIN; WHITE, 2005 ). It is, therefore, a resource available for the writer/speaker to place this/herself interactively in the text regarding his/her rhetorical objectives. In this sense, a text would be the result of the interaction with several other texts and voices, an attempt to describe the author’s stance (his/her position regarding something) at the moment the text is produced.

Finally, graduation might be a means of intensification or mitigation of meanings instantiated in other systems. This system has two mechanisms, force – quantity or intensity of the evaluation – and the focus – prototyping and precision. As Lima-Lopes and Vian Jr. (2007) observe, the focus can be defined in terms of more specific categories, operating taxonomies which specify the participation; force, on the other hand, must cover meanings that can be quantified or intensified.

In this article, my focus is on the system of attitude. This choice is a result of the nature of the comments, which seem to actively judge the candidates, sparing no criticism to their political and personal lives, or highly praise them for their political platform or courage to face the many Brazilian issues. There are some studies in Brazilian Portuguese that discuss how user comments in different social media might be an essential resource for social representation. Lima-Lopes & Pimenta (2017) discuss the sexism in Brazilian football and observe that women always have their knowledge about sports put to the test, compared to men, who are not tested. Women are questioned continuously; they have to prove they can understand the rules and know the historical matches and results. Men, on the other hand, seem to have in their gender a natural acknowledgement of football expertise. Lima-Lopes (2017b) joins register Analysis and Network Science to understand social and informational roles in a WhatsApp community dedicated to homebrew crafting. His results show that the position each person has within the network reflects the role taken in the community. In other words, there is a relationship between the degree of centrality of a participant and the kind of information he (only men were active in the group) offers and demands.

Gabardo & Lima-Lopes (2018) discuss how an Argentinian collective pro-women’s rights was able to mobilise people in four different Latin American countries (Brazil included) and one in Europe (Spain) in some political rallies. The collective was successful in not identifying an agenda of women’s rights with feminist movements and promoting political pressure for the approval of anti-femicide and pro-abortion laws. Finally, Lima-Lopes (2018) reflected on the replication of conservative hate language and non-verified news on Facebook. The author applies some network concepts to the study of collocates, discussing how hate language organises itself in terms of Field and Tenor.

In this article, I have a similar approach by observing which verbs, adjectives and nouns relate more frequently to each candidate. I intend to understand how the users in their comments portray each candidate.

Methodology

The data for this research consists of comments on the interviews of four Brazilian presidential candidates, two males and two females. In order to guarantee that the comments were not made under misinformative/fake pieces of videos, all data was scraped from the official Roda Viva’s YouTube channel.

Comments were scraped seven days after the interview and are in Brazilian Portuguese (BP). The tool used for data collection was YouTube Data tools,² which made it possible to collect all comments from the videos. The videos are still available at Roda Viva ’s YouTube Channel,³ and data scraping is possible, but the dynamic nature of social media might have caused some changes in terms of content.

YouTube Data tools access YouTube API (application programming interface) and scrape video comments and metadata. In order to download the comments, it should be necessary to indicate to the tool the video ID, usually represented by the last few digits in the video’s address. The tool delivers UTF-8 *.tab separated table files, all easily opened in standard spreadsheets. As each comment comes in a single line, it is possible to analyse them separately (regardless of extension). It is possible to know if a comment is original or a reply, the date and time it was posted, how many times the video or an individual comment was “(dis)liked” or “favorited” and who was the author of each comment. All these metadata will possibly be used in a next study, due to this research scope.

Data were processed in an online corpus management tool, Sketch Engine⁴ . Sketch Engine offers a set of tools for corpus analysis such as collocates, lexical collocations comparison, concordance, automatic POS tagging, wordlists and keyword lists extraction, occurrence plot and many others. For this study, the collocates and concordancing functions were used.

The most relevant collocates were calculated using mutual information score ( BREZINA, 2018 ), disregarding prepositions, articles and connectives. Sketch Engine tags corpora with POS information and the node’s syntactic function in relation to a collocate, so it was possible to observe which transitivity relationship the candidates had when collocating with a process. The number of collocates analysed for each candidate is a result of mutual information significance scores (p ≥ 0.05) for each set of comments: 51 for Ciro, 33 for Manuela and 46 for Bolsonaro, 42 for Marina.

There were few interventions in the corpus, mostly to identify the candidates in comments in which they were represented as ele/ela (she/he/him/her) or a nickname (Bolsonaro, for example, is commonly referred to as captain). The reference was written between squared brackets, and the original reference was not changed. The final list of collocates was classified according to the appraisal and transitivity frameworks.

Collocates were later processed in the Gephi ( BASTIAN; HEYMANN; JACOMY, 2009 ), a network plotter able to plot networks of any kind. Gephi also performs several statistical calculations, such as centrality and an in/out-degree ( SCOTT, 2013 ). In this study, the size of each node represents the relevance of the candidate connection to a particular lexical item. Gephi files should follow a particular syntax. Two files should be written, the first file declares the network relationships while the second declares the name and label for each node. The files might be written in an ordinary spreadsheet and then saved as *.csv. Examples of such syntax are on tables 3 and 4 .

Table 3
– Gephi source files (Edges)

Source	Target	Type	Weight
W1	W2	Undirected	34
W1	W5	Undirected	50

Source:

Table 4
– Gephi source files (Nodes)

Label	Word	Type
Manuela	W1	Candidate
Comunista	W2	Adjective

Source:

The words present in each graph had their concordances studied, such concordances provided insights for qualitative analysis. Other researchers have based their research in this methodology, such as Lima-Lopes and Gabardo (2019) , who discussed how women were represented in pro/anti-abortion discourses, Lima-Lopes (2018) , who studied conservative Facebook reactions to an art exhibition, and Lima-Lopes (2020) , who observed how lexical choices were actualised in diachronic football news corpora.

In this paper, examples are in their original BP version without grammar or typos corrections. Below each example, an English translation is present.

Table 5 shows the size of the corpus studied. It is important to note that YouTube users spontaneously generate the number of words; there is no control or selection regarding comment size. The corpus is over two million words, counting only original comments, it brings all the comments present in each video at the moment data was collected, and it represents the full set of data available.

Table 5
– Corpus size

Candidate	Tokens	Types
Marina Silva	58.881	50.055
Ciro Gomes	682.377	579.904
Jair Bolsonaro	757.828	639.552
Manuela D’Ávila	774.858	659.624
Total	2.273.944	1.929.135

Source:

The numbers on table 2 show a very close number of types and tokens. Such numbers seem to be a result of the nature of the texts this study has analysed. Other studies ( LIMA-LOPES, 2018 ) show that there is a strong tendency for large portions of texts to repeat in a social media interaction, a result of three common practises in internet hate language. Firstly, such repetition might be an expression that one belongs to a particular ideological/social group, as users recognise people they share ideological views in the open internet by common lexical choices when discussing a specific topic. So, users tend to copy/paste some portion of texts only adding a few extra words to express some particular point of view. Secondly, although YouTube does not limit the number of characters one might write as do some other platforms like Twitter, these comments tend to be only a few words long, probably a result of the attention span necessary to read them. Finally, internet robots (or bots) are responsible for a great deal of repetition in the open internet, as they tend to repeat the same words over and over.

Regardless of this tendency, no study has been conducted in order to discuss the average extension of YouTube comments and other media (like Facebook). Despite its importance, this would extrapolate the scope of this article.

Table 5 also shows that the number of types and tokens in Marina Silva’s video is much smaller than the other candidates. This difference might be a result of Silva’s status during the time the interviews were conducted, since she was both the candidate with the least numbers in all pools and the one running for the smallest party of all (Rede), by the time the election took place, Marina was amongst the candidates that received the least number of votes.

Final algorithm of data collection and analysis

Finally, this study followed the algorithm of data collection and analysis described below.

• 1.Choose videos from Roda Viva’ s official YouTube channel;

• 2.Scrape data, using YouTube Data tools;

• 3.Upload data to Sketch Engine;

• 4.Calculate collocates;

• 5.Download lists;

• 6.Process Gephi plots;

• 7.Concordancing;

• ● Using Sketch Engine;

• 8.Interpret the results.

• 9.Create tables according to collocates polarities

Results: Jair Bolsonaro

As figure 1 shows, Jair Bolsonaro collocates more frequently with processes in which the candidate plays the modified entity. In Material processes, he is always the Goal in propositions in which he is portrayed as under attack by left-wing journalists (examples 1, 2 and 3). In all cases, he has been ascribed the role of a victim of biased and hypocritical professionals (Ex.: 1 and 2), who have the clear intention to prevent him from winning the elections. In all cases, these journalists play the role of an actor, while Bolsonaro is the victim.

Figure 1
– Collocations of Jair Bolsonaro scores (p ≥ 0.05, n = 46)

Ex.: 1 Esses jornalistas de esquerda, completamente tendenciosos, tentaram encurralar Bolsonaro (…)

Ex.: 2 (…) Os bobocas queriam derrubar [Bolsonaro] e ajudaram e muito (…)

Ex.: 3 (…) Vergonha nacional esses hipócritas atacando o Bolsonaro (…)

Ex.: 4 (…) perguntas estas mais voltadas para tentar denegrir do que saber.

Victimising Bolsonaro is a common strategy in these comments, associating any question that is a little harder as a threat to him as a person and to his political project. As a result, the journalists are considered aggressors and have not only their professional abilities challenged (examples 1 and 7) but are also personally offended (examples 2 and 3). In verbal processes, this pattern is actualised by instantiating the journalists as sayers in propositions that Bolsonaro plays the target (examples 4 and 5). When present, the verbiage is the element that disqualifies Bolsonaro (examples 4 and 5) or a suggestion of questions that should have been made (example 7). Example 8 brings another kind of criticism, now addressed to the YouTube platform, which is accused of suppressing comments in favour of Bolsonaro. This sort of conspiracy theory seems to try to bind people together by a feeling of persecution.

Example 7 is especially important because it brings a definition of what a political interview should be for these commenters: the opportunity for Bolsonaro to deliver his plans for the country. Such a monologue should happen without any questioning or challenging of his political platform or ideological affiliation. It is essential to observe that the law which regulates Brazilian elections demands that each candidate make available in writing for the general public his/her political platform, or plano de governo as we say in Brazil. As the comments seem to show, they had not had contact with this document before the interview and relied entirely on it to know more about Bolsonaro’s plans.

Ex.: 5 #Folhadedesaopaulo lixo tentando acusar o cara [Bolsonaro] de coisas que ele não tem culpa e fazendo perguntas idiotas (…)

Ex.: 6 Que gente cretina...em vez de perguntarem [ao Bolsonaro] o que ele vai fazer de fato por nosso país!!!

Ex.: 7 Achei que os jornalistas do programa Rodada viva fossem mais preparados .…só tentaram desqualificar Bolsonaro (…)

Ex.: 8 Estão falando que estão apagando os comentários a favor neh?

When Bolsonaro plays an active role in the processes, it is possible to observe the projection of some crucial roles in clauses that represent positive evaluations of Bolsonaro and his campaign. In example 9, he is the president responsible for taking Brazil to the future, something possible not only as a result of him being the actor in a material process but also by the presence of a circumstance of behalf. This circumstance represents a common Brazilian expression regarding future positivity, something that belongs to the XXI century is advanced and modern.

Ex.: 9 Bolsonaro está se candiando para governar no século XXI

Ex.: 10 Parabéns Bolsonaro estamos juntos PRESIDENTE 2018

Ex.: 11 Bolsonaro está tranquilo e respondendo com firmeza, sem ser grosseiro.

Ex.: 12 A candidatura de Bolsonaro está crescendo

Example 10 shows a user congratulating Bolsonaro and using a relational process of attribution to project the idea s/he is accompanying him in his journey to the presidency. In this kind of clause, Bolsonaro always takes the role of carrier followed by a circumstance of accompaniment. His performance is also evaluated (ex. 11) positively. He plays either the carrier of a positive attribute (mostly an adjective) or a sayer ( respondendo , ex. 11) in a verbal process modified by a circumstance that carries an optimistic assessment. In the case of verbal processes, a circumstance of mode ( com firmeza and sem ser grosseiro, ex. 11) qualifies his actions as positives. There is also an acknowledgement of his candidature growth through material processes in which it plays the actor (ex. 12).

In these examples, progressive tenses are quite common as they express the simultaneity between two events in Brazilian Portuguese. In some cases, such tenses express the co-occurrence between writing the comment, Roda Viva’s transmission and the user’s interpretation of his performance (ex. 11). His candidature is identified as a path for Brazilian development, mostly because of the relationship between the behavioural and the circumstance of direction (ex. 9). The growth of his candidature (ex. 12) is also characterised as simultaneous to the moment the comments are posted.

Although the number of processes is higher than the nouns and adjectives, it is relevant to notice that the latter represent more central and relevant connections with Bolsonaro (see figure 1 ). This relevance is a result of the node size and their proximity to the central node, which represents the candidate. Two of these relationships seem to call the reader’s attention: despreparado (unprepared) and futuro (future), see figure 1 . These two words highlight the two different evaluations of Bolsonaro as a person and as a politician. Despreparado is used both as an attribute in relational processes and as an epithet. In the first case (examples 13 and 14), the candidate plays the carrier as the attribute is either intensified by an adverb ( visivelmente in example 13) or is part of a list of insults (example 14), which intensify its meaning by the accumulation of negative prosody. Epithets tag along with the candidate in non-relational clauses (example 15). Usually, it is part of an evaluation of the candidate’s performance and lack of good qualities for getting into the office.

Ex.: 13 O Bolsonaro é visivelmente despreparado para o cargo.

Ex.: 14 Bolsonaro é um populista desmiolado, limítrofe e despreparado .

Ex.: 15 Que candidato despreparado , não responde uma pergunta, é este que quer ser presidente do Brasil?

Most of the adjectives are disqualifying Bolsonaro. The comments here tend to project a negative evaluation of his personal and political life. The strategy is to associate him with different sorts of prejudice, most of them pinpointed from previous interviews or public speeches. They are present in propositions which either compare Bolsonaro to Donald Trump (ex. 16) and associates negative characteristics to both or enumerate some negative adjectives (ex. 17 and 18). In the first case, the intention is to demerit his association with the US president by ascribing similar attitudes towards political minorities to both of them. A common strategy is to place adjectives one after the other sequentially. Such a strategy results in a “pile” of characteristics that seems to intensify Bolsonaro’s adverse judgement. It can be described in Brazilian Portuguese as creating a cumulative intensification of the criticism.

The only adjective that collocates exclusively in positive oriented evaluations is bom (good , ex. 19). It occurs in a formulaic expression in Portuguese, something similar to You’d better or You are better in standard English. There is a tone of irony due to the similar prosody of Jair (first name of Bolsonaro) and já ir (already going to), projecting his victory and the need for the opposition to adjust.

Ex.: 16 Bolsonaro é o Trump Brasileiro Homofóbico e Racista .

Ex.: 17 Despreparado, ineficiente, hipócrita e mentiroso. Machista , misógino, homofóbico e racista.

Ex.: 18 Que idiota gente! Racista e homofóbico !

Ex.: 19 É bom Jair se acostumando.

Nouns (see figure 1 ), on the other hand, insatiate mostly mixed evaluations when it comes to the candidate. Futuro , for example, comes in contexts that project Bolsonaro as either the next president of Brazil or criticise the show’s participants for not making questions which project the future of the country (in this case with him as president). On example 20, he is modified by an identifying relational process, in which he is the token, and the value is the projection of his election. In example 21, the same projection takes place, but the talk show is now accused of trying to change people’s votes. The user is reacting towards the journalists’ performance since they were hard on the candidate. In this context, any question that does not favour the candidate is considered an aggression. The natural ambiguity of the word future in Portuguese, which might mean both “future” and “next” might lead us to see it in (ex. 22) as a possible reference to the candidate who is representative of a new era.

The word família (family) is also representative of a common thought during Bolsonaro’s campaign. The best part of Bolsonaro discourse features the idea that equal rights to political and social minorities, mostly women, gays and trans people, will destroy the traditional heterosexual families, which are in the rightness of the Lord. Clauses like (ex. 23) are the result of such a way of thinking: Bolsonaro is the “battle” for defending the right virtues, all who have equality and human rights as core values are leftists and must be fought. This approach ascribes to Bolsonaro a messianic mission, as we can read in (ex. 24).

Ex.: 20 Esse é nosso futuro Presidente!

Ex.: 21 Roda Viva, mas vocês falharam. Bolsonaro será o nosso futuro presidente (…)

Ex.: 22 São perguntas desnecessárias, precisamos saber do futuro e não do passado.

Ex.: 23 Bolsonaro presidente para o bem da família tradicional o terror dos esquerdistas (…)

Ex.: 24 Ninguém supera meu herói Bolsonaro, que nada o abale meu salvador (…)

In this messianic context, Bolsonaro plays the role of token in relational identifying processes. The values are nouns which characterise him as Brazil’s saviour (example 24). This sort of evaluation is also present in clauses that actualise other nouns in the role of value, such as solução (solution, ex. 25) and opção (option, ex. 26), and fenômeno (phenomenon, ex. 27).

Nouns are also usually present in examples that criticise Bolsonaro, and they also work by associating the candidate as the Carrier/Token in Relational processes, as it reads on examples like 28 and 29. They mostly try to disassociate the idea that he is the best option, or even that he knows what he is going to do. This idea is related to the image that the candidate seems to project both inside and outside Brazil. Circumstantial elements are essential in this context since they create the relationship between the local and the global (ex. 28), in addition to projecting possible external ideological interferences (ex. 29).

Ex.: 25 O Brasil está a deriva Bolsonaro é a solução para o nosso país.

Ex.: 26 (…) a melhor opção que vejo no momento é Jair Bolsonaro.

Ex.: 27 Essa é mais uma estratégia pra camuflar o fenômeno que está sendo o bolsonaro!

Ex.: 28 Bolsonaro é a vergonha do Brasil no Brasil e no mundo.

Ex.: 29 BOLSONARO É UM BONECO DA NOVA ORDEM MUNDIAL.

Bolsonaro is a blockhead in the new world order

Table 6 brings a summary of the frequency of types and tokens represented in figure 1 . Although it was possible to identify some heavy criticism towards Jair Bolsonaro, the general tone of the comments is positive. The most common strategy is to evaluate the journalists and other candidates negatively, to project a positive image towards Bolsonaro. In this context, positive evaluation is mostly indirect, creating favourable prosody. This is why the classification of lexis ( table 6 ) represents what I should call a reverse strategy: the negative evaluation of Bolsonaro’s adversaries is a way to increase his popularity in discursive terms. Every time we read a lexical item criticising a journalist or a Politician, a positive picture of him is drawn. Consequently, there is a reverse polarity in which a negative evaluation of one turns into a positive evaluation of the other. Similar results were observed by Lima-Lopes, Mercuri and Gabardo (2020) , who study how commenters react to fact checking agencies posts on Facebook.

Table 6
– Bolsonaro’s collocates according to polarity

Polarity	Types	Tokens	%
Positive	26	710	34.3
Negative	13	55	2.66
Neutral	4	1306	63.04
Totals	46	2070	100

Source:

Ciro Gomes

As figure 2 reads, verbal processes are in clauses that portray two polarised evaluations regarding Gomes. Chamar (to call,⁵ as a free translation) is responsible for defending Gomes from some different kinds of criticism. Example 30 shows Ciro as a target in a verbal process, in which the criticism plays the role of verbiage. The strategy here is to show that Ciro is honest because he refused to get some financial advantages granted by the Brazilian political system in a positive verbal action, introduced in a projected clause. The function of this clause is to disqualify such criticism by representing him as a sayer in a verbal process ( recusou , declined).

Figure 2
– Collocations of Ciro Gomes (p ≥ 0.05, n = 51)

Example 31 makes a comparison between Ciro Gomes and Jair Bolsonaro, who is not present in the proposition but can be inferred by the adjective nazi . Ciro plays the target in propositions which the verbiage insults him. He is associated with the word coronel ( colonel ); a common insult in the Brazilian political context. However, both Ciro and Bolsonaro are Carriers in propositions in with users qualifies a colonel ⁶ as better quality than a Nazi , a characteristic commonly associated with Bolsonaro. Ciro plays the target and carrier while the attributes tend to question his character in private life (ex. 32).

Ex.: 30 Quem chama o Ciro de mentiroso quer aparecer às custas do único político que se recusou a receber 3 aposentadorias (…)

Ex.: 31 Podem chamar o Ciro de Coronel, melhor ser chamado de coronel do que ser chamado de nazista (…)

Ex.: 32 Em carta, prefeito cearense chama Ciro Gomes de vagabundo e usuário de drogas (…)

Ex.: 33 As pessoas criticam o Ciro, mas não vejo ninguém apresentar argumentos (…)

Criticar (to criticise, example 33) brings positive meanings, mostly by identifying the criticisms as not correctly supported by arguments. Ciro plays the target, while the supportive proposition is in a parataxis relationship. Apoiar (to support), which is a verbal process in Brazilian Portuguese due to its declarative characteristics, is present in propositions that also instantiate “if” clauses. Such conditionals tend to be a means to criticise both supporters and no-supporters of Ciro, who is a target in these processes. On example 34, non-supporters are sayers in propositions that project an insulting clause. The insult instantiates in a circumstance of place. On example 35, supporters are carriers in relational processes which represent them as carriers of mental diseases.

Ciro Gomes regularly instantiates the range (range or beneficiary (ex. 38)) also in behavioural clauses. As a range, Ciro is usually subject to direct attacks by the users, reporting the discourse of external sources such as journalists ( ex. 36), or other users (ex. 37). External sources usually are held responsible for the information.

Users are accused of promoting racism as they attack Ciro due to his origin, ex. 37. Such propositions are a piece of evidence of two things. Firstly, there is tension between some political groups in the North and South of Brazil; secondly, they reveal a partial knowledge of Ciro’s real background. Even though he has had all his political career in Ceará, a state in Northeast, he was born in Pindamonhangaba in São Paulo, a state in the Southeast.

Ex.: 34 Se não apoiarem o Ciro, vão pra lata do lixo completamente.

Ex.: 35 Se você é contra governo de esquerda e apoia o Ciro, você precisa ser estudado ⁷ (…)

Ex.: 36 Assisti o vídeo do Jornalista_01; desmascara o Ciro com muito humor.

Ex.: 37 Não adianta atacar o Ciro por ele ser Nordestino (…)

Ex.: 38 Quem pesquisa Vota Ciro Gomes! (…)

Other critical behavioural processes are eleger (to elect, ex.39), ouvir (to listen, ex. 40) and conhecer (to meet, ex. 42). In both processes, Ciro plays a range in propositions that ascribe him the role of a political saviour. In some cases, he is the carrier of a relational attributive process which relates to a behavioural clause by parataxis, example 39. Example 40 brings the same relationship but a significant difference: here, the carrier/attribute pair is represented by two different sentences. The first represents Ciro and his opponents, while the second acknowledges the other candidate as intellectually inferior.

Halliday (1978 , 1994 ) takes English as the basis for the development of SFL. Although both English and Portuguese are Indo-European languages, some structural and semantic differences between them might lead to a review of some elements within the theory. Halliday (1994) would define behavioural processes as mental actions that would have some material outcome. This would be the criterion, for example, to distinguish processes like see/hear (involuntary physical activities) and watch/listen (physical actions with a mental purpose/intention). However, Halliday and Matthiessen (2014) point out that such processes are the least distinct kind because they do not have clearly defined characteristics; they would be partly like material partly like mental.

It is relevant, though, to keep in mind that Halliday’s (1994) definition places behavioural processes in-between material and mental processes in both semantic and grammatical terms. This is so because mental processes in English are rarely expressed in progressive form, while the material ones are. As behaviours are halfway between these two, they are representations of our inner world which take material expressions and, therefore, may or may not take progressive forms.

Nevertheless, this difference seems to be a little bit fuzzier in BP, mostly because both mental and material processes can take progressive form; the difference is, therefore, entirely semantic ( LIMA-LOPES, 2001 ). It opens a window to interpret processes like eleger (ex.39) and ouvir (ex.40) as material and conhecer (ex.42) as both mental or material depending on the context (translating as meet or know). If on the one hand Eleger in a sense expressed in ex.39 would represent a coordinated action of voting, which could be material if we consider the corporeal displacement and typing components involved — in Brazil elections use electronic ballot boxes —, on the other, it would also represent a conscious choice for Ciro, a mental action which would result from considering his political platform. One could think that ouvir (ex.40) and conhecer (ex.42) would be similar to eleger. Ouvir represents not only the action of listening — or hearing since both might be exchangeable in some contexts in BP — which would be material, but also comparing and evaluating what Ciro and the other candidates have said — what represents a mental component. Conhecer (ex.42) also entails both the mental action of considering Ciro’s political project, but also the material action to come across him in the interview.

On account of these processes’ semantic ambiguity, my choice is to classify them as behavioural. However, I am not assuming that Halliday’s idea of a halfway meaning would work flawlessly in BP. In BP, the mental component of material processes is something that makes itself present intensively, which would make the idea of a ‘least defined’ element somewhat ambiguous. Behavioural processes, therefore, could be defined in BP as complex processes, processes which are not in-between but are both mental and material at the same time.

Examples 41 and 42 bring a more confessional tone since both clauses identify Ciro as the reason for their hope for Brazil and the elections; behavioural processes are represented in hypotactic clauses that indicate it as the cause of the glimpse of a better future. A very intricate grammatical system instantiates this meaning since Ciro is both range and carrier simultaneously in a single clause complex. It makes the evaluation spread across all the complexes.

Ex.: 39 Temos que eleger o Ciro, ele é a solução que o Brasil precisa (…)

Ex.: 40 Quando você vê e ouve CIRO GOMES e os demais candidatos é observado um abismo intelectual!

Ex.: 41 Já tinha desistido de votar, mas voltei a me sentir esperançosa depois que passei a ouvir o Ciro falar (…)

Ex.: 42 (…) tinha desistido do Brasil até conhecer ⁸ o Ciro!

The nouns (25.45%) ascribe to the candidate negative evaluations. The most common pattern is the association with negative nouns such as lixo (garbage, ex. 43) and farsa (farse, ex. 44), or vira-casaca (ex. 45), which represents a straightforward criticism, commonly without a clear justification. As Lima-Lopes (2018) points out, this sort of aggression seems to be common in Brazilian ideological discussions. The author observes that people tend to take political discussions to the personal level, most of the time repeating offensive discourse without observing possible consequences.

Some propositions compare Ciro to former presidents of Brazil, mostly in a very negative way. In this strategy, the user tries to show that Lula, Dilma, and Ciro are the same political persona with cosmetic differences. The comparison between Lula and Ciro tends to be offensive. They rely mostly on Lula’s educational background (ex. 47), he does not hold a college degree as Ciro does, or on physical aspects (ex.46), Lula had a finger amputated during his time as a steelworker.

The comparison with Dilma tends to be related to a single fact: she is a woman (ex.48). The argument takes a traditional Brazilian chauvinist saying by which two people are virtually the same, but with a single difference: their gender. In such a framework, people would say a man (x) is a woman(y) in his pants (when a man is compared to a woman), while the woman (y) is a man (x) in her skirts (when a woman is compared to a man).

Ex.: 43 Ciro é outro lixo como tem trouxa pra votar nesse cara?

Ex.: 44 Ciro é uma farsa !

Ex.: 45 Ciro é um socialista vira-casaca (…)

Ex.: 46 Ciro é um Lula com 10 dedos e sem moral (…)

Ex.: 47 Ciro é um Lula com diploma (…)

Ex.: 48 Ciro é uma Dilma de Calça (…)

Adjectives are the least frequent collocations with Ciro (21,63%) and bring mixed evaluations in clauses where he is the carrier in relational processes. The negative judgement is frequent, and it is related to his political positions and his proximity to Lula, to whom he served as a minister. He is characterised as igual (just like, example 49) Lula, as desprezível (despicable, example 50) or negatively characterised as socialista (socialist, ex. 51) and comunista (communist, ex. 52).

Ex.: 49 Ciro é igual a Lula!

Ex.: 50 Ciro é desprezível por suas posições ideológicas (favorável aos corruptos Lula e Maduro).

Ex.: 51 Mas o Ciro é socialista , de modo que perderíamos ainda mais liberdades (…)

Ex.: 52 Ciro é comunista declarado (…)

Ex.: 53 CangaCiro, você é um engodo , um embusteiro e mentiroso .

The only two positive adjectives are claro (clear, ex. 54) and inteligente (intelligent, ex. 55), and they are related to the candidate’s capacity of being clear in his answers.

Ex.: 54 A frase de Ciro é clara e a fonte dos dados inquestionável (…)

Ex.: 55 Ciro é inteligente e complexo.

Another common offence is CangaCiro (ex . 53). Without a possible translation into English, CangaCiro is a neologism which takes the word cangaceiro , a participant of an outlaw gang in the Brazilian Northeast semiarid during the late 1800s and early 1900s, and the name Ciro (Canga+Ciro). It is a result of a certain phonetic similarity; the term exposes prejudice against Ciro’s political origin, the State of Ceará. As example (ex. 53) shows us, its use is related to some personal offences.

Table 7 brings a summary of the frequency of types and tokens represented in figure 2 . Ciro Gomes is also more positively evaluated. However, his numbers are much closer than Bolsonaro numbers, while the positive evaluation tends to repeat the lexicon a little bit less than the negative. While the first is more frequent and presents a richer vocabulary, the latter tends to reiterate the same idea constantly. The higher number of the neutral lexicon is related to delexicalised verbs like fazer (do or make) and ter (have) commonly instantiating auxiliary functions. The criticism towards Ciro Gomes tends to be a result of prejudice against his political origin.

Table 7
– Gomes’ collocates according to polarity

Polarity	Types	Tokens	%
Positive	28	334	44.18
Negative	15	229	30.29
Neutral	7	193	25.53
Totals	50	756	100

Source:

Manuela D’Ávila

Manuela collocates ( figure 3 ) mostly by behavioural processes, probably a result of personal manifestations regarding the context in which the interview took place. According to the comments, Manuela was interrupted much more frequently than her male counterparts, and the acknowledgement of such interruptions was the most common topic in the discussion (examples 56 to 59). In such processes, she plays the range and sometimes the active role in the projected verbal process (ex. 56). However, it is vital to observe she is not victimised as Bolsonaro is. Most comments acknowledge such a situation and criticise the talk show for that (examples 58 and 59). In some cases, intensification is present in a circumstance of frequency (examples 57 and 58) or in a relational attributive sentence (example 59). Users try to be neutral, acknowledging the interruption issue, but not supporting her politically (ex. 60) in propositions Manuela plays the role of behaviour and sayer in projected verbal processes.

Figure 3
– Collocations of Manuela D’Ávila (p ≥ 0.05, n = 33)

Manuela is the target in verbal (ex.: 61 and 62) and phenomenon in mental processes (ex.: 63 and 64). Positive evaluations always come embedded into relational attributive clauses which single out who Manuela’s primary opponents are in terms of their profession or social class (ex.: 61). The criticism is directed towards her supporters, who are identified as sick (ex.: 62). The use of essa (this, ex.: 62) is very significant in BP, because it might associate an idea of inferiority and contempt to Manuela.

Ex.: 56 Não deixaram Manuela falar (…)

Ex.: 57 (…) interromperam a Manuela mais de 60 vezes (…)

Ex.: 58 Acabam por atacar a Manuela. Lamentável (…)

Ex.: 59 Mas a forma como trataram a Manuela neste programa foi muito desrespeitoso.

Ex.: 60 Não estou defendendo Manuela. Estou defendendo o direito dela de falar.

The mental processes actualise mixed evaluations. Negative clauses criticise Manuela’s supporters by associating to them a knowledge of jumentos’ language (donkeys, ex. 63). Manuela also plays phenomenon in positive evaluations (ex. 64). Mental sentences are in a parataxis relationship with relational attributive clauses in which are an epithet expressing optimistic feelings.

Ex.: 61 Todos que criticam a Manuela D’Avila são banqueiros (…)

Ex.: 62 (…) como existe pessoas doentes que defendem essa Manuela (…)

Ex.: 63 Não sei a linguagem dos Jumentos. Por isso vc entendeu a Manuela e eu não.

Ex.: 64 Acabei de conhecer a Manuela e fiquei feliz!

Manuela is the sayer in verbal processes, which mostly reflect an evaluation of her interviewers (ex. 66) as well as her performance (ex. 65). The frequent interruptions are again a recurrent topic. However, Manuela is represented positively: she is responsible for answering all questions and being strong (ex. 65) and for discussing the real needs of Brazil (ex. 66). She is also portrayed as someone unable to answer the questions correctly, using her time in the talk show only to get some votes (ex.67).

Ex.: 65 (…) parabéns a Manuela essa mulher de garra respondeu todas as perguntas.

Ex.: 66 Não deixam ela nem falar. eu não entendo manuela não falou nada de ruim ela só falou coisa que o brasil precisa.

Ex.: 67 Relaxa que a manuela se deu bem nessa entrevista, ganhou exposição falando só abobrinha e ficou como vitima.

Dar (give) e ganhar (get or win), ex. 67, would challenge some standard ground process classification in English. Silva, Souza e Souza (2019) discuss the classification of dar based in a semantic taxonomy introduced by Silva (2019) and Rassi and Vale (2013) . According to the authors, dar might extrapolate its material sense, as it would depend on its interaction with the scope to establish its actual meaning. In this sense, the relationship of dar and the noun phrase (scope) defines not only a better part of the semantic function it instantiates but also some grammatical features. As Silva, Souza e Souza (2019) observe, some possible patterns would support their claim in BP: if one considers dar (ex.67) in isolation of bem (well), one would probably classify such a process as a transitional material process. However, a speaker of Portuguese would interpret that D’Ávila has somehow benefited from dar in a way which would be not necessarily material; Dar could be easily probed as a circumstantial relational process ( Manuela foi bem nessa entrevista /Manuela went well…). Dar performs a relational meaning mostly due to its relationship with bem and the circumstance of place ( nessa entrevista /in this interview).

Ganhar seems to be like dar , because most of its meaning depends on the relationship with exposição (directly translated as “exposition”), ex. 67. Mostly, it reflects a strong collocation that brings traces of beneficiary to Manuela, who grammatically plays the actor. In this context, ganhar could also be probed as a relational process .

The adjectives evaluate Manuela D’Ávila’s performance as a candidate (ex. 68) and as an interviewee (ex. 69), as well as her ideological positions (ex. 70). There are also some personal offences directed towards the candidate (ex. 71 and 72). Negative evaluations (ex. 68) are much more common, and they do not bring any discussion on Manuela’s political platform, they seem to be entirely related to her ideological framework. Comments mostly judge Manuela for being either a feminist (ex. 70) or a member of the communist party (ex. 76) would make her less suitable for being in the office. It is essential to notice that this does not also come with any further justification: to be a feminist or a communist is something terrible in itself. Words like comunista (communist) and patética (pathetic) (ex.: 71 and 72) are equal in their negative meaning. Most of the negative adjectives are associated with Manuela without a clear justification.

Manuela is the carrier or possessor in relational clauses which the evaluation comes as an Attribute (ex. 68, 71 and 72) or in a projected clause (ex. 69 and 70). Such projection takes place in either nonverbal (ex. 69) or relational attributive (ex. 70) clauses. Repetition of tonic vowels ( muuuito, ex. 68), uppercase ( Patética , ex. 71), all caps ( CRETINA , ex. 72) and quotation marks (ex. 77) are common resources to demonstrate the intensification.

Ex.: 68 Nossa, a Manuela é muuuito fraca (…)

Ex.: 69 Manuela muito inteligente e paciente com esse bando de estupidos mal educados!

Ex.: 70 A garota tem orgulho de ser feminista , como se isto fosse algo bom (…)

Ex.: 71 Manuela vc é Patética !

Ex.: 72 Manuela vc é CRETINA (…)

Manuela is also commonly portrayed in clauses which either evaluate her physical look or express a chauvinist and sexual content (ex. 73 - 74). In all these examples, Manuela performs the carrier, identified or possessor.

Adjectives like bonita (beautiful) and bonitinha (cute) express judgment of Manuela’s appearance. Bonitinha , in addition to an aesthetic appreciation of her body, also brings the use of suffix *.inha (o), which should express the diminutive of nouns in BP (ex. 74 and 76), which gives it an ironic tone. Such use would sound like she was close to being non-beautiful, or like her beauty would not be enough. Some of these sentences try to describe Manuela in a very offensive way, questioning her sexual behaviour and representing her as a lascivious woman (ex.75).

Ex.: 73 Essa piranha, o que ela tem de bonita tem de inútil essa lata de lixo (…)

Ex.: 74 Bonitinha mas tão burrinha Gostosa!

Ex.: 75 Linda , gostosa , safada, mentirosa, perigosa e sem moral. Adoro esse tipo de mulher. É a mulher certa para uma noite de orgias.

Ex.: 76 Que comunistinha gostosa em gente.

What a hot little communist folks!

In some clauses, there is a direct attempt to be rude and very offensive, sometimes by swearing (ex. 73). It is a common strategy to describe Manuela as incapable of being both beautiful and intelligent (ex. 74). In other cases, the users try to create an accumulation of lascivious characteristics, in order to disqualify her as a possible choice: she is good for sex and nothing else (ex.75). Manuela is also commonly compared to Dilma. The tone is also offensive: Dilma is stupid (ex. 78) and ugly, while Manuela is even more dumb, but more beautiful (ex. 79). In rarer cases, the clauses are equally offensive to both, but Manuela always is described as hotter and younger (ex.: 77 and 79).

Ex.: 77 Esta mulher parece a Dilma , só que “ bonitinha “

Ex.: 78 E eu achando que a burrice da Dilma era insuperável!

Ex.: 79 Essa é uma capivara Dilma mais nova e gostosa (…)

Table 8 brings a summary of the frequency of types and tokens represented in figure 3 . Manuela D’Ávila has an evaluative profile completely different from previous pre-candidates. She is evaluated more negatively than positively, while neutral evaluations are very rare. The results showed that being a female candidate was quite important to this result. She is also the first candidate who had a lexical item which might represent both positive and negative evaluations ( mulher) .

Table 8
– D’Ávila’s collocates according to polarity

Polarity	Types	Tokens	%
Positive	10	173	10.79
Negative	16	575	35.85
Neutral	6	27	1.68
Pos/neg	1	829	51.68
Totals	33	1604	100

Source:

Marina Silva

Marina Silva ( figure 4 ) evaluated both in the political (ex.80) and personal levels (ex. 81, 82, and 83). When positively evaluated, Marina is a range of behavioural processes in which a circumstance instantiates some political identification (ex.80) or character evaluation (ex. 81). A set of characteristics delivered by cumulative list (ex.82) is also a common strategy, which maximises her positive features by creating a cumulative effect in BP.

Figure 4
– Collocations of Marina Silva (p ≥ 0.05, n = 42)

Personal offences are present in clauses Marina plays the carrier in attributive processes, mostly rude words without any possible reason (ex. 84). It is common for the users to say they are not voting for Marina due to her appearance or clothing choices (ex.85), which are, then, evaluated negatively. Marina is portrayed just the opposite to Manuela, the latter is beautiful and hot, while Marina is questioned for her looks all the time.

Ex.: 80 Eu considero a Marina de centro-esquerda (…)

Ex.: 81 Vejo a Marina como uma pessoa equilibrada (…)

Ex.: 82 Mulher forte e genial Marina Silva sempre observei Marina Silva (…)

Ex.: 83 Acho engraçado que as pessoas que ofendem a Marina aqui nos comentários não possui nenhum tipo de argumento.

Ex.: 84 De que raça é essa cachorra chamada Marina (…)

Ex.: 85 Ñ votaria e nem uma presidente de COOOOQUE.

As we can see in figure 4 , Marina instantiates the active participant in fewer processes. In most cases, these processes evaluate her political performance (ex. 88) or her performance as an interviewee (ex. 86) positively. When users discuss her political career, their comments are related to Marina’s opinions on former President Dilma Rousseff (ex. 88), in which Marina instantiates the sayer in verbal processes, or to her projects for 2018 elections (ex. 89), in which she instantiates the behaviour in behavioural processes.

When users discuss her performance as an interviewee, Marina also appears as behaver and sayer. However, she instantiates a different grammatical configuration. The range is implicit, and the target of its verbal processes are the questions she should have answered (ex. 86 and 87).

Ex.: 86 (…) amei esse roda vivaa.. sério, não sou fã da marina mas ela conseguiu dar a volta por cima de todas as alfinetadas com respostas ao nível

Ex.: 87 Marina não falou coisa com coisa

Ex.: 88 Marina disse , na campanha de 2014, que Dilma mentia ao dizer que não sabia da corrupção na Petrobrás (…)

Ex.: 89 A única candidata com dignidade e idéias para governar o Brasil Marina (…)

Marina receives mixed evaluations when she collocates with adjectives and nouns. She performs the carrier in attributive relational processes, while the elements that characterise her play the role of attributes (ex.90). Positive evaluations tend to be presented as a list (ex. 90 and 91), creating a cumulative effect in which a series of positive features sequentially placed. I would like to draw particular attention to the word diva (ex.90) which is used in BP to designate a woman admirable for her intelligence, sensual and seductive power. Although she receives a positive evaluation, Marina seems to be praised for being a woman, not for her political ability. Marina is the carrier in relational processes which delivered (ex. 92) negative evaluations. The presence of hypotaxis or parataxis clauses that intensify such evaluations (ex. 93 - 95) is frequent. In these clauses, Marina is made present by a cohesive element and instantiates one participant in a concessive clause that introduces a flaw of character attributed to her (ex.94).

Marina is a candidate related to conservative Christian bases and such relationship motivate some users to criticise her. Comments judge her negatively for being a member of a centre-left party and for having less-conservative attitudes towards controversial issues such as abortion (ex. 95). In such cases, she instances the possessor in possessive relational processes and the sayer in a paratactical verbal process.

Ex.: 90 Marina é ética , eloquente, inteligente, consciente, pacifica e não passiva, mulher forte, guerreira, pulso firme sim, estudiosa, diva, tudo que precisamos.

Ex.: 91 Mulher é forte e genial Marina Silva

Ex.: 92 Marina é omissa .

Ex.: 93 Marina é tão fraca , que se ela der um peido ela dá 2 passos pra frente!!!

Ex.: 94 Não acho que a Marina seja corrupta, mas quem evita dizer o que pensa, não tem a minha confiança.

Ex.: 95 Não tem certeza das próprias convicções pois se diz cristã evangélica mas defende o aborto.

As Manuela, Marina is also judged for her physical appearance (ex. 93). While Manuela is hot, beautiful and lascivious, Marina’s appearance (ex. 93 and 98) is criticised for not following some beauty standards. Marina is verbally attacked for being too thin (ex. 98), a characteristic which might be an indication of political weakness. Such a framework is always instantiated in uncivil comparisons that aim to humiliate the candidate, who instantiates the carrier in attributive relational processes. In some cases, such negative evaluation is present in adverbial hypotactic sentences, in which Marina can instantiate actors in material or behavers in behavioural processes (ex. 93 and 98).

Unlike all other candidates, Marina is the only one who had comments questioning her racial status (ex. 96 and 97). The issue seems to be beyond the simple identification of her origins. This representation becomes a social judgement with an air of irony as one could infer some disdain for the possible identification of Marina as an Afro-Brazilian citizen (ex. 96). In these comments, being of a mixed-race may seem more valuable and noble than being black, since these users see her characterisation as Afro-Brazilian as disrespectful (ex. 97). Such a denial is related to culture that values the need to identify Marina as non-black.

Ex.: 96 A Marina é negra ? Nem tinha percebido.

Ex.: 97 Marina Silva não é negra , ela é mestiça! ela vem da Amazônia do norte, respeitem a mestiçagem brasileira!

Ex.: 98 Marina Silva vc não governa nem teu quintal. Nada sabe, um aspecto horrível , uma cara de SONGA MONGA.

Ex.: 99 Marina é tão safada quantos as abortistas

Ex.: 100 COMUNISTA VAGABUNDA . HERPES DA POLÍTICA: SÓ APARECE A CADA QUATRO ANOS, QUANDO A IMUNIDADE ABAIXA.

Ex.: 101 Marina é esquerdista ferrenha, ativista ambiental e, como tal, sua missão é tirar das pessoas o direito à propriedade.

Ex.: 102 marina saiu do PT mas tem propostas iguais as do PT

Ex.: 103 DESLIKE DESLIKE DESLIKE DESLIKE DESLIKE DESLIKE DESLIKE DESLIKE DESLIKE

Personal offences are common in the corpus, and they take multiple forms. Marina is treated as a safada (whore, ex.:99 and 100) or as a disease for having a less conservative position regarding the abortion or for taking part in the Brazilian presidential elections (herpes, ex. 100). Other common criticisms reflect a total misunderstanding of Marina’s political platform, as long as she has never positioned against the right of private property in the interview (ex.101), or her relationship with PT, which she was part of during the most prolonged period of her career (ex.102). Finally, example (103) brings another common list effect: the word deslike (a typo of dislike) is repeated 786 times in sequence in each of three different comments. The general idea is to express disagreement with Marina’s answers during Roda Viva. However, no explanation or discussion is present; there is merely the direct aggression represented by the all caps repetition.

Table 9 brings a summary of the frequency of types and tokens represented in figure 4 . Results show that Marina is the pre-candidate with the least number of positive evaluations. This tends to be a result of the constant repetition of “DESLIKE” a typo of “dislike” that seems to be posted insistently throughout the comments (ex. 103). Although this study did not aim at discussing whether the commenters are human or internet robots (bots), the repetition of a single lexical item seems to indicate it was automatically posted. If so, it brings a new topic to discuss in further studies, how to consider automatically posted material.

Table 9
– Silva’s collocates according to polarity

Polarity	Types	Tokens	%
Positive	10	14	93.69
Negative	12	906	4.86
Neutral	19	47	1.45
Totals	41	967	100

Source:

Final remarks

This article aimed at discussing the representations of four pre-candidates for the presidency of Brazil through a survey of comments related to their interviews in the official channel of a top-rated Brazilian talk show, Roda Viva . My objective was to observe whether the two most commented male candidates, Ciro Gomes and Jair Bolsonaro, would have different representations when compared to only the two female candidates interviewed, Marina Silva and Manuela D’Ávila. The theoretical foundation was on the SFL, more specifically in Transitivity and Appraisal Systems. The SFL was responsible for providing the textual analysis framework needed to understand the ideological principles underlying lexicogrammatical choices.

The methodology included internet data scraping, besides its processing through the graph visualisation software and an online concordancing. Despite the quantitative procedures of the data collection, the analysis was also qualitative and focused on the survey of the processes along with evaluative lexicon.

The results seem to show that male and female candidates are represented differently. Jair Bolsonaro is represented either as a victim or as the saviour of the country. He seems to represent the traditional family and values. The comments showed that his audience expected to watch a monologue in which he would hold an exposition about his plans for the country. Most of the adjectives are disqualifying Bolsonaro and some comments project a negative evaluation of his political life.

Ciro Gomes is represented as the best option due to his academic knowledge and his political experience. On many occasions, he brings hope and convinces people who had given up voting, not doing so. Ciro is criticised for being associated with leftist parties and for their political origin, the Brazilian Northeast. Sometimes the question of origin is used as a personal offence with demonstrations of racism.

Some comments on Manuela D’Ávila are related to the nature of the interruptions and the progress of her interview. There is no debate about her plans for the country, and her political origin is a reason for personal offences. Although some comments praise her performance in the interview and her intelligence, most of them are comparing her to other leftist politicians, such as Dilma, often through sexist remarks. In these comments, Manuela is young, hot and lascivious, despite being communist and dumb.

Some comments highlight the suitability of Marina Silva for presidential office, although most of them associate her to the left in a pejorative way. Her performance in the interview is questioned continuously, with few positive evaluations. Her physical appearance is also an issue but in a different way. Marina is questioned for her aesthetic choices and for not following a pre-established beauty pattern. Some rude and extremely offensive criticisms are addressed to her personally. Marina is the only candidate who is questioned about her race, sometimes ironically. Her possible identification as an Afro-Brazilian is, inclusively, seen as an offence by some users.

Although some comments praise Manuela D’Ávila and Marina Silva, they are not messiahs or saviours of Brazil. These results show that both male and female candidates are constantly criticised for their political origins and affiliation. All candidates are also evaluated according to their capacity of governing Brazil. Such criticism is sometimes harsh and direct. However, Jair Bolsonaro and Ciro Gomes are never questioned about their looks, their race as they are not the target of any kind of uncivil sexual comments. Marina Silva and Manuela D’Ávila, on the other hand, are constantly judged in these terms.

Acknowledgements

I would like to thank FAPESP (2016/11230-5) for financial support for this research.

REFERÊNCIAS

• ANDROUTSOPOULOS, J.; TEREICK, J. YouTube. In: GEORGAKOPOULOU, A.; SPILIOTI, T. (ed.) The Routledge handbook of language and digital communication . Oxon: Routledge, 2015. p.354–370.
• BAKER, P. (ed.). Contemporary corpus Linguistics . London: Continuum, 2009.
• BAKER, P.; MCENERY, T. (ed.) Corpora and discourse studies: integrating discourse and corpora. Houndmills; New York: Palgrave Macmillan, 2015.
• BARABÁSI, A.-L. Linked: the new science of networks. Cambridge: Perseus, 2002.
• BASTIAN, M.; HEYMANN, S.; JACOMY, M. Gephi: an open source software for exploring and manipulating networks. In: INTERNATIONAL AAAI CONFERENCE ON WEBLOGS AND SOCIAL MEDIA, 3., 2009, Califórnia. Proceedings [...]. California: Association for the Advancement of Artificial Intelligence, 2009. p.361–362.
• BREZINA, V. Statistics in corpus linguistics: a practical guide. Cambridge; New York: Cambridge University Press, 2018.
• COELHO, L. M.; BAPTISTA, M. A história da inserção política da mulher no Brasil: uma trajetória do espaço privado ao público. Revista Psicologia Política , São Paulo, v.9, n.17, p.85–99, 2009.
• DELU, Z. Role relationships and their realisation in mood and modality. Text , Berlim, v.11, n.2, p.289–318, 1991.
• EGGINS, S.; MARTIN, J. R. Genres and registers of discourse. In: VAN DIJK, T. A. (ed.). Discourse as structure and process: discourse studies: a multidisciplinary introduction. London: SAGE Publications, 1997. p.230–256.
• FLORENTINO, K. Mulheres na política: a importância de sua representatividade. Politize , [ s. l .], 18 out. 2018. Disponível em: https://www.politize.com.br/mulheres-na-politica/ Acesso em: 22 jan. 2019
  » https://www.politize.com.br/mulheres-na-politica/
• FOX, J.; CRUZ, C.; LEE, J. Y. Perpetuating online sexism offline: anonymity, interactivity, and the effects of sexist hashtags on social media. Computers in Human Behavior , Oxford, v.52, p.436–442, nov. 2015.
• GABARDO, M.; LIMA-LOPES, R. E. de. Ni una menos: ciência das redes e análise de um coletivo feminista. Humanidades & Inovação , Palmas, v.5, n.3, p.44–58, 2018.
• GUERRA, R. Na política brasileira, 15 pioneiras desde o início do século XX até a Presidência. O Globo , [ s. l. ], 3 mar. 2016. Disponível em: https://acervo.oglobo.globo.com/em-destaque/na-politica-brasileira-15-pioneiras-desde-inicio-do-seculo-xx-ate-presidencia-18824935 Acesso em: 23 jan. 2019.
  » https://acervo.oglobo.globo.com/em-destaque/na-politica-brasileira-15-pioneiras-desde-inicio-do-seculo-xx-ate-presidencia-18824935
• HALLIDAY, M. A. K An introduction to functional grammar . London: Arnold, 2004.
• HALLIDAY, M. A. K. An introduction to functional grammar . London: Arnold, 1994.
• HALLIDAY, M. A. K. Language as social semiotic: the social interpretation of language and meaning. London: University Park Press, 1978.
• HALLIDAY, M. A. K.; HASAN, R. Language, context and text: aspects of language in a social-semiotic perspective. Oxford: Oxford University Press, 1991.
• HALLIDAY, M. A. K.; MATTHIESSEN, C. M. I. M. Halliday’s introduction to functional grammar . 4.ed. Oxon: Routledge, 2014.
• HALLIDAY, M. A. K.; MATTHIESSEN, C. M. I. M. Construing experience through meaning: a language-based approach to cognition. London; New York: Continuum, 2004.
• HALPERN, D.; GIBBS, J. Social media as a catalyst for online deliberation? exploring the affordances of Facebook and YouTube for political expression. Computers in Human Behavior , Oxford, v.29, n.3, p.1159–1168, 1 maio 2013.
• HUNSTON, S.; THOMPSON, G. Evaluation: an introduction. In: THOMPSON, G.; HUNSTON, S. (ed.). Evaluation in text: authorial stance and the construction of discourse. Oxford: Oxford University Press, 2000. p.1–25.
• IEDEMA, R. Multimodality, resemiotization: extending the analysis of discourse as multi-semiotic practice. Visual communication , London, v.2, n.1, p.29–57, 2003.
• KRESS, G. Multimodality: a social semiotic approach to contemporary communication. London: Routledge, 2010.
• LIMA-LOPES, R. E. de. Immigration and the context of Brexit: collocate network and multidimensional frameworks applied to appraisal in SFL . Muitas Vozes , Ponta Grossa, v.9, n.1, p.410-441, 2020.
• LIMA-LOPES, R. E. de. O conservadorismo como ideologia: contribuições da ciência das redes para a linguística sistêmico funcional. Letras , Santa Maria, v.28, n.56, p.43–69, 2018.
• LIMA-LOPES, R. E. de. Reflexões sobre as possíveis contribuições da linguística do corpus para a gramática sistêmico funcional: transitividade e classificação de processos. Caletroscópio , Ouro Preto, v.5, n.9, p.9–25, 2017a.
• LIMA-LOPES, R. E. de. Análise de registro e ciência das redes estudando um grupo de whatsapp dedicado à produção de cerveja artesanal. Hipertextus: Revista Digital, Recife, v.16, p.134–161, 2017b.
• LIMA-LOPES, R. E. de. Transitivity in Brazilian Greenpeace’s electronic bulletins. Revista Brasileira de Linguística Aplicada , Belo Horizonte, v.14, n.2, p.413–439, 2014.
• LIMA-LOPES, R. E. de. Processos relacionais em cartas publicitárias. Revista Brasileira de Linguística Aplicada , Belo Horizonte, v.8, n.1, p.35–69, 2008.
• LIMA-LOPES, R. E. de. Estudos de transitividade em língua portuguesa: o perfil do gênero cartas de venda. 2001. Dissertação (Mestrado em Linguística e Estudos da Linguagem) – Pontifícia Universidade Católica de São Paulo, São Paulo, 2001.
• LIMA-LOPES, R. E. de; GABARDO, M. Ni una menos: A luta pelos direitos das mulheres na Argentina e suas representações no Facebook. Revista Brasileira de Linguística Aplicada , Belo Horizonte, v.19, n.4, p.801–824, 2019.
• LIMA-LOPES, R. E. de; MERCURI, K. T.; GABARDO, M. Avaliatividade em comentários sobre postagens dedicadas à verificação de notícias falsas nas eleições presidenciais de 2018. Cadernos de Linguística e Teoria da Literatura , Belo Horizonte, v.1, n.4, p.1–25, 2020.
• LIMA-LOPES, R. E. de; PIMENTA, I. #Mulheresnofutebol: transitividade e avaliatividade na identificação de padrões sexistas. Humanidades & Inovação , Palmas, v.4, n.6, p.116–132, 2017.
• LIMA-LOPES, R. E. de; VIAN JUNIOR, O. Resenha de The Language of evaluation: appraisal in English. DELTA: Documentação de Estudos em Lingüística Teórica e Aplicada, São Paulo, v.23, n.2, p.371–381, 2007. Resenha de: MARTIN, J. R.; WHITE, P. R. R. The Language of evaluation: appraisal in English. Hampshire: Palgrave MacMillan, 2005.
• MARTIN, J. R. Meaning matters: a short history of systemic functional linguistics. Word , Philadelphia, v.62, n.1, p.35–58, 2016.
• MARTIN, J. R.; WHITE, P. R. R. The language of evaluation: appraisal in English. New York: Palgrave Macmillan, 2005.
• MOOR, P. J.; HEUVELMAN, A.; VERLEUR, R. Flaming on YouTube. Computers in Human Behavior, Online Interactivity: Role of Technology in Behavior Change, Oxford, v.26, n.6, p.1536–1546, 1 nov. 2010.
• PINTO, C. R. J. Paradoxos da participação política da mulher no Brasil. Revista USP , São Paulo, v.49, p.98–112, 2001.
• RASSI, A. P.; VALE, O. A. Tipologia das construções verbais em português do Brasil: uma proposta de classificação do verbo dar. CALIGRAMA , Belo Horizonte, v.18, n.2, p.105-130, 2013.
• ROCHA LIMA, C. H. DA. Gramática normativa da língua portuguesa Edição revista segundo o novo acordo ortográfico ed. Rio de Janeiro: José Olympio, 2010.
• ROSSI, M. Brasil, a lanterna no ranking de participação de mulheres na política. El País , São Paulo, 31 mar. 2018. Disponível em: https://brasil.elpais.com/brasil/2018/03/27/politica/1522181037_867961.html Acesso em: 22 jan. 2019.
  » https://brasil.elpais.com/brasil/2018/03/27/politica/1522181037_867961.html
• SANTOS, B. C. dos. Política em números: 5 dados sobre participação das mulheres na política brasileira. Politize !, [ s. l .], 17 mar. 2017. Disponível em: https://www.politize.com.br/participacao-das-mulheres-na-politica-brasileira/ Acesso em: 22 jan. 2019
  » https://www.politize.com.br/participacao-das-mulheres-na-politica-brasileira/
• SCOTT, J. What is social network analysis? London; New York: Bloomsbury Academic, 2013.
• SILVA, J. J. D.Um estudo sistêmico-funcional de orações com o verbo dar no português brasileiro 2019. 241f. Tese (Doutorado em Letras) – Universidade Federal de Pernambuco, Recife, 2019.
• SILVA, J. J. D. da; SOUZA, M. M. de; SOUZA, A. A. de. Sistematização das orações com o verbo dar no português brasileiro. Intercâmbio: Revista do Programa de Estudos Pós-Graduados em Linguística Aplicada e Estudos da Linguagem, São Paulo, v.42, p.49–67, 17 dez. 2019.
• TABOADA, M. Sentiment analysis: an overview from linguistics. Annual Review of Linguistics, Palo Alto, v.2, n.1, p.325-347, 2016.
• STUBBS, M. British traditions in text analysis: firth, halliday and sinclair. STUBBS, M. In: Text and corpus analysis . London: Blackwell, 1996. p.23–50.
• TABOADA, M. Sentiment analysis: an overview from Linguistics. Annual Review of Linguistics , Palo Alto,v.2, n.1, p.325–347, 14 jan. 2016.
• LIMA-LOPES, R. E. de; MERCURI, K. T.; GABARDO, M. Avaliatividade em comentários sobre postagens dedicadas à verificação de notícias falsas nas eleições presidenciais de 2018. Cadernos de Linguística
• TABOADA, M.; GRIEVE, J. Analysing appraisal automatically. In: AAAI SPRING SYMPOSIUM ON EXPLORING ATTITUDE AND AFFECT IN TEXT: THEORIES AND APPLICATIONS, 2004, Califórnia. Proceedings [...]. Califórnia: Association for the Advancement of Artificial Intelligence, 2004. p.158-161.
• THOMPSON, G. Introducing functional grammar . London: Arnold: Oxford University Press, 2004.
• THOMPSON, G. Introducing functional grammar . London: Arnold, 1996.
• THOMPSON, G.; ALBA-JUEZ, L. (ed.) Evaluation in context . Amsterdam: John Benjamins, 2014.
• THOMPSON, G.; THETELA, P. The sound of one hand clapping: the management of interaction in written discourse. Text , Berlim, v.15, n.1, p.103–127, 1995.
• VIAN JUNIOR, O.; LIMA-LOPES, R. E. de. A perspectiva teleológica de Martin para a análise dos gêneros textuais. In: MEURER, J.; MOTTA-ROTH, D.; BONINI, A. (ed.) Gêneros: teorias, métodos, debates. Campinas: Parábola, 2005. p.29-45.

• 1
  At the time, there was another female candidate Vera Lúcia, who was not interviewed. TV Cultura justified her absence by stating that Vera Lúcia did not fit the criteria for candidate selection for interviews. Vera Lúcia pointed out that, although she tried to contact Roda Viva production several times, she received no answer and was not invited to the show. Later in the campaign, Manuela D’Ávila quit her own pre-candidature and joined the ticket with Fernando Haddad as his Vice-President.
• 2
  Available at https://tools.digitalmethods.net/netvizz/youtube/
• 3
  Available at https://www.youtube.com/user/rodaviva
• 4
  Available at https://www.sketchengine.eu/
• 5
  There is not a direct equivalence to chamar in English, a translation might highly depend on the context, as we can see at examples 30, 31 and 32.
• 6
  In Brazilian Portuguese, besides the military rank, Colonel might also refer to traditional and powerful landowners in the northeast, which were very powerful and behaved most like feudal masters. Initially, coronelismo (as a political and social movement) lasts until the early 1950s, however, the term is still used to characterise influential politicians with electoral feuds in the northeastern states.
• 7
  In Brazilian Portuguese the expression “ você precisa ser estudado ” is similar to comparing someone to a mental freak, someone who needs to be studied by a psychiatrist given his/her incoherent mental state.
• 8
  In Brazilian Portuguese, conhecer might mean both know and meet. The translation choice justifies as the user seem to have had his/her first contact with Ciro Gomes

Publication Dates

• Publication in this collection
  09 Mar 2022
• Date of issue
  2022

History

• Received
  17 July 2020
• Accepted
  22 Oct 2020