Disintermediation and disinformation as a political strategy: use of AI to analyse fake news as Trump’s rhetorical resource on Twitter

The communicative effects of disintermediation caused by social media promote the expansion of personalist and emotional political discourses that reach the audience directly and evade the traditional journalistic filter. This phenomenon leads to new political communication tactics, but also exposes citizens to potentially fraudulent, contaminated or pola-rised content. In this context, framed in post-truth, the term ‘fake news’ gains relevance as a way of referring to disin-formation and as a political and performative argument that can be weaponised. This research aims to analyse such use in the discourse of the former president Donald Trump during his presidential term (2017-2021), focussing on Twitter as the main platform in his political communication strategy online. To analyse this, we resort to a methodological triangulation of content, discourse


Introduction
The crystallisation of social media implies a qualitative leap in online communication through applications that prioritise interpersonal contact ( Van-Dijck, 2016), as well as massive, multidirectional interaction and participation (Beer;Burrows, 2007;Pérez-Salazar, 2011;Pfister, 2011).Under these digital structures, those surfing the internet become potential prosumers who fulfil the simultaneous roles of producers and consumers (Ritzer;Dean;Jurgenson, 2012).Thus, they can select, distribute, and curate the information on different platforms (Hernández-Serrano et al., 2017;Timoshenko;Hauser, 2019) without the need for specialised knowledge (Muswede, 2022).
The appearance of the user as a participant in the flow of information (Pavlicková; Kleut, 2016), and the competition between their own content and that offered by the media, whose distribution is measured, in part, by the automation of the algorithms, the search engines and the viral flow (Lokot;Diakopoulos, 2015), counter the intermediary role of the journalists as the traditional gatekeepers, since the "power to inform and transmit reality no longer belongs solely to the media" (López-Jiménez, 2014, p. 84).
These mechanisms and forms of participation contribute to the disintermediation of mass communication in the digital environment, in which the people "reject the representative mediation and prefer the self-representation of their particular, subjective reality" (Mencarelli, 2021, n.p.).
Faced with this change in the communicative process, the sources of information are also forced to adapt their position: politicians, enterprises and groups with differing power and social influence become prosumers (Berrocal et al., 2014).They modify their communicative strategies and areas of diffusion through these networks, in order to launch their messages directly to the potential users, voters or consumers.It is a strategy that takes advantage of micro-targeting as a form of aiming at audience niches through the exhaustive study of their habits or preferences, so as to reach them in the most direct and personalised way possible (Barbu, 2014).It is a practice embedded within the marketing techniques in which artificial intelligence (AI) plays a fundamental role, and which allows massive quantities of data to be processed (Kotras, 2020) in order to personalise the content (Milan, 2015;Sánchez-García et al., 2023).This gives rise to such paradoxes as the personalisation of the masses through decisions taken by machines (Ritzer, 2015) that are also capable of dealing with the public discourse (Gillespie, 2010).This mechanisation affects equally both the automation of political information (Bradshaw et al., 2020) and its influence on public opinion (Murthy et al., 2016); this is due to its effect on journalistic objectivity (Carlson, 2018;Wu, 2019) and its relationship with fake news (Zimmer et al., 2019).
In this context, there occurs the transition from mediated politics (Castells, 2009), which needed the media as intermediaries with public opinion, to disintermediated politics, in which leaders and entities transmit their message directly to their audience through social media.Thus, the persuasive skills of the politicians are now being developed through a connected, telematic media flow (Muswede, 2022) that provides electronic devices with political goals (Grossman, 1995) and gives the Internet direct democracy (Morris, 2001(Morris, , p. 1033)).This is an environment that can favour interactivity, participation, and cooperation (Coleman, 2005;Fernández-Castrillo, 2014); yet it can also bring dangers, such as the encouragement of 'popular politics' (Berrocal et al., 2022) which treats leaders as celebrities; the opportunistic use of citizens as viral instruments (García-Orosa, 2021, p. 3); or the impulse towards a new Caesarism (Rubio-Fabián, 2019, p. 46) that can pose a threat to democracy (Borgesius et al., 2018).

Disinformation and fake news in political discourse: the case of Donald Trump
Disintermediated political communication provides an opportunity for the persuasive strategies of leaders, but it also puts reception at risk, as the content is prone to manipulation and may reach public opinion "contaminated by the wishes of the person producing it" (Enguix, 2020, p. 26).The selective presentation of social media foments intrinsic effects of this environment, such as filter bubbles (Pariser, 2011) or echo chambers (Cinelli et al., 2021), both intensified by the use of such mass disinformation tools as the creation of social bots to automatically disseminate fraudulent contents (Boshmaf et al., 2011) and to "potentially manipulate discussions in social media (…), creating false narratives that take root in a large percentage of the population" (Rossetti; Zaman, 2023, n.p.).
It is a context that leads to the polarisation of both politics and the people, and which fosters an environment that favours sensationalism, polarisation and disinformation via these platforms, as opposed to professional journalism in crisis (Cano-Orón et al., 2021a), as has been studied in such countries as Germany (Papakyriakopoulos et al., 2018), Russia (Ribero et al., 2019), Spain (Cano-Orón et al., 2021b), India, or the United Kingdom (Cheng, 2019), as well as the electoral processes in the UK and the USA (Fincham, 2019).More particularly, the figure of the USA's ex-president, Donald Trump, has generated numerous works of research (Raynauld;Turcotte, 2018;Singh;Wijegunawardhana, 2018).
His direct communication through Twitter (now X) has been a watershed in political communication, using it beyond the institutional, unlike his predecessor Obama, stating his political opinions and judgements without intermediaries.It was a propagandistic and rhetorical wager that gave him an advantage over his rivals (Muswede, 2022;Das-Sarma, 2016), which he also endowed with a controversial viral and showy resonance (Gómez-García et al., 2019).
"The unique combination of the power provided by his position as a world leader and the extremely colloquial tone of his publications meant his tweets had a great «success».At the end of 2020 he had accumulated almost 80 million followers and he had published over 50,000 tweets before his Twitter account was permanently suspended in January of 2021" (Machus et al., 2022, n.p.).
As part of this political strategy based on direct communication, the accusations of informative falsehood against his opponents, including politicians, spokespersons, and the media, represent a discursive constant associated with the context of post-truth favoured by the increase in fake news (Journell, 2017;Lorenzo;Manfredi, 2019;Bleakley, 2018).A series of attacks used reiteratively by Trump and reproduced by like-minded sources as figures of authority "exalted" his public and generated an effect of illusory truth that distorted their perception (Froehlich, 2020).
Although there are several concepts to refer to different types of false or manipulated information (Carmi et al., 2020), "the term 'fake news' refers to a whole range of information types, from honest errors of little impact and satirical content to high impact manipulative techniques and malicious inventions" (Kapantai et al., 2020, p. 5).
It is a term that is overloaded (Wardle; Derekshan, 2017) and inadequate for describing the complexity of the phenomenon of disinformation in an era already earmarked as "the era of fake news" (Albright, 2017) and which, nevertheless, acquired "worldwide relevance in 2016, during the presidential elections in the USA (…).The term was widely used (or abused) to characterise almost any content that entered into conflict with the points of view or the programme of a particular party" (Kapantai et al., 2020, p. 2).
The concept, associated with other previously existing concepts, such as post-truth, not only took on relevance because of its meaning, but also because of its instrumentalisation or performative impact (Farkas;Schou, 2020, p. 6) as a political weapon.Donald Trump intentionally focused part of his discourse on this, starting "a war of rhetoric against the established media, labelling them as the fake news media" (Farkas;Schou, 2020, p. 6).
In this light, the above studies not only consider the interested propagation of fake news or the use of bots to support his discourse (Rossetti;Zaman, 2023), but also the Trumpist narrative of painting as fake the information he considered to be detrimental by means of a discursive deflection (Ross;Ribers, 2018).This practice is considered to be an instrumentalisation and has recently been investigated in the American context with respect to -its reception by the public opinion (Tong et al., 2020); -the influence of biased informative sources in the success of this strategy (Froehlich, 2020;Meirick;Franklyn, 2022); -the satirical content or memes (Smith, 2019); -its use to discredit the established media (Rossini et al., 2021;Happer;Hoskins;Merrin, 2018); and -the effects of its dissemination on the audience's informative perception (Tamul et al., 2019).
Despite the academic attention dedicated to the effects of disinformation, "there has been less research into how the political elites and the experts use fake news as a weapon to discredit the media" (Rossini et al., 2021, p. 676).
Keeping in mind the above-described context, this research offers a dual analysis, thematic agenda, and emotions, of the rhetorical and political use of the term fake news by the ex-president of the USA, Donald Trump, in his discourse published in social media during his complete term of office (2017)(2018)(2019)(2020)(2021).In particular, the object of the study focuses on Twitter as the principal platform used in his disintermediated discursive strategy (Machus; Mestel; Theissen, 2022) with two main aims: a) to analyse the thematic agenda, the policies, and the players (the media, cabinet, and opposition) that the ex-president refers to when using the term fake news; b) to establish the emotional level or the polarity (positive, negative, or neutral) of his discourse based on experimentation with artificial intelligence (AI) tools using machine learning systems.
Disintermediated politics can promote participation and direct democracy, but also the instrumentalization of technologies, personalism or spectacularization Both of them allow us to perceive a panoramic view of the instrumentalisation of the term fake news as part of the disintermediated discursive strategy and the disinformation practiced by the ex-president of the USA.At the same time, they also allow us to test technological AI tools that facilitate an automation process in the case of the digital discourse analysis.
Starting from these objectives, this research proposes three main hypotheses: -H1.Donald Trump instrumentalises the term "fake news" with a rhetorical use.
-H2.The ex-president of the USA uses the label "fake news" as an ambivalent argument to discredit the sources that oppose his discursive line with a triple strategy of attack-defence-imposition of ideas.-H3.The messages in which Trump has recourse to the term "fake" against information from the established media are characterised by a certain polarity aimed at discrediting the said media so the people will no longer trust them.

Materials and methods
This research uses a triangular methodology, based on content analysis, of the discourse and emotions, centred on two axes: thematic agenda (E1) and emotional level or polarity (E2).The sample is made up of the tweets published by the ex-president with the term "fake news" during his mandate (2017-2021).
In a first phase, the sample is obtained from the tweets in "The Trump Archive": https://www.thetrumparchive.com It is an open-access platform that brings together all the messages emitted by the politician from the very start of the records (December 10 th 2016) until the account was suspended on January 8 th 2021 (Courty, 2021), and which has already been used in previous works of research (Quealy, 2021;Magallón, 2018;Meeks, 2019).
Starting from this open database, a first filtration is carried out based on tweets from when he was sworn in on January 20 th 2017 until the elimination of his profile that fit the search term (N=970).The sample was then further filtered with the elimination of retweets, deleted tweets and those that could not be visualised or codified due to a lack of context, leaving a final sample (N=768) downloaded from The Trump Archive in JSON (see Annex 1) format which allows them to be codified using a "software" of our own elaboration that facilitates the collection and automated data analysis in the empirical work.
In a second phase, this same filtered sample is submitted to an analysis of emotions through the techniques of Lexicon and Deep Learning.Figure 1 summarises the technical procedure of each of the proposed axes, and which will be considered individually in greater detail in the following subsections.
The design work and testing of the tool is distributed according to the areas of specialisation of each author of this research work: the programmer developed and implemented the application according to the needs marked out by the researchers, who chose the labels and variables and carried out the testing and codification.

Thematic agenda: content and discourse analysis (Axis 1)
The thematic analysis of Axis 1 was carried out through a content and discourse analysis, two methodological techniques that allow rationales to be established by combining categories (Piñuel-Raigada, 2002) and textual structures to be analysed ( Van-Dijk, 1990).The aim is to identify in which themes Donald Trump uses the term "fake news" as a political argument in his Twitter discourse.These were selected on the basis of a prior sample analysis, detecting the most repeated ones and grouping them by similarity (Márquez- Domínguez et al., 2017).This allows them to be fitted into four main categories and 52 multiple choice variables (see Annex 1), since a tweet can be about one or more topics.
The absence of filters in online political communication exposes citizens to potentially fraudulent, contaminated or polarized content This analysis is made operational through a web application of our own elaboration based on the Angular framework to visualise and categorise the tweets (Figure 2), obtained from The Trump Archive; and in the administrator of the mon-goDB database to consult, analyse and add information.This allows us to save the codification and interrelate the final results on the basis of consultations in the database.During the said coding process, a first pre-test of the intercoders was done following the criteria of Krippendorf (2004) concerning a subsample of 20% (N=153).This managed to achieve an average agreement of 93.5%.This manual phase of the content analysis was complemented by an automated sentiment analysis.

Polarity: sentiment analysis with AI: using Lexicon and Deep Learning (Axis 2)
This second phase offers a double automated sentiment analysis, "a popular method used to analyse discourse through the identification of the valence in the data from the text" (Misiejuk et al., 2021, p. 376).It is a semantically oriented tool that helps to understand "public opinion concerning controversial questions in a more accurate, complete and accessible way " (Abdulla et al., 2013, s.p.).It has also already been applied to the study of political discourse in Twitter (Kaur et al., 2021).In the case of Trump, this technique has been used in previous studies applied to concrete aspects of his discourse, such as the economy (Machus; Mestel; Theissen, 2022; Colonescu, 2018), the general elections (Chandra; Saini, 2021; Xia; Yue; Liu, 2021), or domestic policy matters such as the management of the coronavirus pandemic (Dwianto; Nurmandi; Salahudin, 2021).
An analysis is used that is focused on the emotional level through a polarity qualified as positive-negative-neutral. We do so through the combination of two AI methods: Lexicon through the use of a dictionary; and a pre-trained machine learning model based on Deep Learning and Natural Language Processing (NLP).The combination of these two techniques achieves "more reliable and valid" results (Wahleed et al., 2021, p. 100).
For the first type of analysis focused on the polarity, Lexicon is used as an unsupervised technique in which, in this case, a dictionary is used; that is to say, a list of words with an already assigned numerical polarity.In this case, the Node.js and the Natural code library are used.The latter includes the dictionary Afinn-165, which "assigns values to emotions ranging from -5 to 5; this allows a more precise quantification of the emotional content of the words than other similar lexicons" (Colonescu, 2018, p. 379).
The automated analysis commences with the filtering and cleaning of the sample (N=768).Hyperlinks and special characters (punctuation signs, hyphens, symbols, etc.) are eliminated, verbal formulas with apostrophes are replaced in or- der to maintain the structure of the data (e.g., "you're" becomes "you are"), and words that do not provide any information, such as conjunctions, determiners or numerals are also eliminated, thus increasing the accuracy of the analysis as much as possible.
Having once achieved the definitive sample, the analysis of emotions divides each unit of analysis or tweet into words, which are then assigned a value individually, generating a mean polarity (Abdulla et al., 2013).In order to classify the results, the research divides the range of values into Very Positive, Positive, Neutral, Negative or Very Negative. 4 The second procedure combines Google Natural Language and Amazon Comprehend, both commercial tools that analyse emotions based on Deep learning and NLP, with two of the best scores in accuracy from the available tools (Ermakova; Henke; Fabian, 2021).The technique is unsupervised and aimed at the polarity; it is based on an intelligent, trained model, which allows a better accuracy and performance (Dang; Moreno-García; De-la-Prieta, 2020).To do so, we used the sample of tweets already refined using Lexicon.A programmed analytical instruction was executed for each of the individual tweets in the libraries available on each platform, obtaining the result with the polarity data, which was then dealt with using a summation algorithm to group the scores and thus obtain the relative percentage of each polarity within the sample.

Results
The analysis of content, discourse and emotions illustrates how and to which topics Donald Trump directed the concept of fake news as a political argument in his online discourse, as well as an approximation of the level of emotion or polarity.The following subsections deal with the results of the thematic content analysis and the data from the sentiment analysis, in that order.

The fake news agenda of Donald Trump
The thematic agenda in the ex-president's discourse on fake news (Figure 3) shows a preponderance of governmental policies and actions both within the country and internationally (46.87%), as well as specifically concerning the Trump Administration (32.9%).Beyond the general accusation of fake news, which the ex-president used to defend the said policies, he takes direct aim at the media in over half his messages (52.6%).The last great block is taken up by his political opposition, principally the Democrat Party (20.7%), even though he tends to label the media as "the real opposition party."

Political self-referral (C1) and supporting his cabinet (C2)
The detailed results of the first category of analysis concerning the contents that refer to Trump's policies and cabinet reveal a personalist agenda, in which his own figure as president stands out, occupying 24.7% of the mentions (Figure 4).Self-referral in his discourse mainly concerns defending himself through accusations aimed at the media and the opposition, which he characterises as haters.Worth noting is a high level of personalism, with such phrases as "your favourite president", "a true champion of civil rights" or "your all-time favourite duly elected President, me!"This result is further confirmed through the analysis of the messages concerning the electoral campaign (16.3%),where he aggrandises his own figure using positive polling results, to which he refers as the "the polls that matter", as opposed to the "fake news suppression polls" that do not show him as the winner.This praise of his own person can also be seen in other aspects of his political administration, when he attributes to himself the good handling of foreign policy (22.5%): "there has never been a president who has been tougher (but fair) on China or Russia"; positive numbers concerning the economy and employment (6.38%): "my Administration and I built the greatest economy in history (…) saved millions of lives"; or for his handling of immigration (3.77%).Fake news arises as a concept associated with misinformation, but is also used in a post-truth context as an argument or discursive deviation The discursive tactic of the former president reveals a use of his institutional figure for the sake of political personalization and polarization in the absence of filters or journalistic verification

Relation with the opposition: politicians (C2) and discrediting of the media (C3)
The ex-president relates his discourse on fake news to the political opposition, but also to the media that are not like-minded (Figure 5).In the first case (20.7%), he refers to the opposition in mainly a generalised way, alluding to the Democrat Party as a whole (13.2%).As for particular candidates, Trump focuses on Joe Biden (4.6%) as his main adversary in the following General Elections of 2020.To a lesser extent and during the first half of his term of office, he also refers to Obama (2.2%) and Hillary Clinton (2%) as leaders of the previous opposition.Trump dedicates derogatory expressions and epithets to all of them: "democrats of the radical left who know nothing," "crooked Hillary Clinton," "slow Joe," "Obamagate," "illegal democrat witch hunt." On the other hand, his allusions to the media, either in general (19.5%) or aimed at particular targets (33.1%) are present in more than half the messages published by Trump with respect to fake news (52.6%).He uses the term indistinctly to refer to false news and the media itself, which he calls "the real opposition party" and directly links them to the Democrat Party by using such expressions as "the "fake news" media," defining them as their "partner in crime," "lap dog," "their vehicle," or the "illegal democrat/fa news media partnership."In the same way, he criticises the use of unrevealed sources and encourages his audience not to believe such news (Figure 6): "The most often used phrase in the Lamestream Media, by far, is 'sources say', or 'officials who spoke on the condition of anonymity' (…) which allows Fake News to make up a phony quote from a person who doesn't even exist".We have detected more references to the audiovisual media than to the written media and the social media.In the first case, Trump focuses on CNN (13.9%), and in particular, the ex-presenter Chris Cuomo.Similarly, he considers NBC (6.9%),ABC (2.7%) and CBS (1.3%) to be opposition media.His relationship with Fox News (5.1%), however, changes over time.
In the case of the written media, he focuses on The New York Times (7.7%) and The Washington Post (4.8%).Allusions to social media (3.8%) indicate the relevance he gives them in spreading his discourse outside of the traditional media: "I use social media not because I like it, but because it is the only way to fight against a VERY dishonest and unfair 'press,' now often referred to as the Fake News Media;" "Fake News Media coverage of me is negative, with numerous forced retractions of false stories.Hence my use of social media, the only way to get the truth out." This position changes when Twitter blocks profiles or like-minded messages: "they and the fake news, working together, want to silence the truth;" "Twitter is interfering in the 2020 Presidential Election".

Emotions and polarity in the discourse of Donald Trump
The results of the sentiment analysis allow us to qualify the polarity of Trump's messages in his digital narrative.The combination of the two methods (dictionaries and machine learning) and the three tools (Afinn, GNL and Amazon Comprehend) allow us to obtain detailed comparisons of the contents of his discourse.Figure 7 offers two types of data: on the left, the complete results of each tool in accordance with the particular variables offered by each one; and on the right, the results set out under three variables (negative, positive, or neutral), in which the values of Very Negative to Negative and Very Positive to Positive from GNL and Afinn are grouped so as to be able to compare them directly with Amazon Comprehend, which adds its own variable of Mixed.
As a whole, we can observe a mainly negative polarity in all three services, oscillating between 65% and 84%; a minority of positive between 26% and 8%; and a residual of neutrality, with 14% being the highest.These results demonstrate Trump's hostile tone which, combined with the previous thematic analysis, he uses both to attack others and to defend himself.It is a discourse with a negative tone that comes out to his global audience, directly, without filters.
The interrelation between the emotions analysis and the codified thematic categories in the content analysis reflects concrete disparities with respect to the negativity expressed by Donald Trump in his digital discourse.The figure is lower when referring to his policies (with an average polarity index of -0.0878) and cabinet (-0.0594), but increases his negative emotional load in the case of his political opposition (-0.116) and, in particular, the media (-0.1776).Finally, the breakdown of the tweets into words (eliminating conjunctions, determiners, and verbs) allows us to identify which are the most frequent in his discourse, confirming the previously found result (Figure 8).Worth noting are the references to the media, either directly (CNN, TNYT, MSDNC, NBC) or using related terminology (source, press, stories, fact, report).Other negative, polarised adjectives of note that can further help us to understand how Trump frames his discourse with respect to fake news include hoax, false, radical, corrupt, dishonest, phony, or pathetic.

Discussion and conclusions
This research, focused on analysing the political use of the term "fake news" as a rhetorical resource in the discourse of Donald Trump in Twitter during his term of office (2017-2021), allows us to examine his thematic agenda and to outline the tone and polarity of his disintermediated strategic narrative through a content analysis using a software that digitalises the manual codification, alongside the testing and application of machine learning for the automated emotions analysis.
Both objectives lead us to confirm the first hypothesis, which proposed that Donald Trump instrumentalises the allusion to fake news as a rhetorical resource (H1).Thus, Trump focuses his strategy on defending his measures in both home and foreign affairs through attacks and even verbal aggression, using the label of fake to discredit the sources that contradict him (Farkas;Schou, 2020;Karpantai et al., 2020), while defending himself at the same time.It is an ambivalent narrative that ratifies the second of our hypotheses (H2), in which he uses a mainly negative tone (72%), in particular aimed at certain figures of the public agenda, such as journalists or politicians, who he insults or gives nicknames to as a method of mockery, thus accentuating the polarisation factor.To be precise, Trump alludes to the not like-minded media, such as CNN, The New York Times, or The Washington Post (Rossini et al., 2021;Happer; Hoskins; Mirrin, 2019), linking them with the opposition as being in an alliance or a witch hunt, in an attempt to discredit them, therefore confirming the last hypothesis (H3).At the same time, he seizes the opportunity to heap praises upon himself and his cabinet; a more personal than institutional position that coincides with the tactics of political personalisation and turning leaders into celebrities (Muswede, 2022;Das-Sarma, 2016;Berrocal et al., 2022).
On the whole, the evidence shows a performative or rhetorical use of the term "fake news"; as either a "discursive diversion" resource (Ross; Ribers, 2018) or as a political weapon (Farkas;Schou, 2020).The evidence also shows an instrumentalisation of the technological devices (Grossman, 1995), in this case, social media, as platforms for generating direct echoes without the verification or intervention of the journalistic filter (Lokot;Diakopoulos, 2015).All this allows him to contaminate the information himself (Enguix, 2020), as well as to take advantage of the automated action of bots to spread and manipulate the online dialogue in order to influence users' perceptions (Boshmaf et al., 2011;Rossetti;Zaman, 2023, n.p.).
Although this research is limited by the fact that the analysis partly depends on automated tools that are constantly developing and which can still improve their reliability indices, protocols, or identification of bias, among other aspects (Sadia et al., 2018), the results can be considered valid and of interest for characterising Trump's digital narrative concerning fake news from a discursive and content perspective as a whole.A performative and instrumentalising strategy of the term that, related to disintermediation and disinformation in political communication, requires constant studies applied to new discourse and political campaigns.It is a concept that, although widely studied, continues to expand, and evolve in the rhetorical and literary academic spheres, concerning both causes and cases, as well as effects.
In this sense, the application of AI tools in the methodology and the testing of software to analyse content can be considered useful for studies in social sciences that, as in the case of communication in general or political communication in particular, focus on the analysis of content and discourse.The application permits the digitalisation of this process and the gathering of data that can be adapted to any study of this sphere based on the collection and labelling of units of analysis of the media and social media.Similarly, the contribution of the dataset can be considered of value for consulting the sample and reusing it in other prospective works.

Figure 1 .
Figure 1.Technical procedure to analyse the content, discourse and emotion using own elaboration software and AI techniques

Figure 2 .
Figure 2. Visualisation of the own elaboration APP to analyse the thematic agenda of the tweets of Donald Trump concerning fake news

Figure 4 .
Figure 4. Details of Donald Trump's thematic fake news agenda with respect to his policies and cabinet

Figure 6 .Figure 7 .
Figure 6.Tweet from Donald Trump accusing journalists of being "corrupt" and of falsifying their sources.Source: The Trump Archive

Figure 8 .
Figure 8.Most frequent words concerning fake news in the discourse of Donald Trump in Twitter