With the rise of social media, cyberbullying and online spread of hate have become serious problems with devastating consequences. Mentimeter is an interactive presentation tool enabling the presentation audience to participate by typing their own answers to questions asked by the presenter. As the Mentimeter product is commonly used in schools, there is a need to have a strong toxicity detection program that filters out offensive and profane language. This thesis focuses on the topics of text pre-processing and robustness to datashift within the problem domain of toxicity detection for English text. Initially, it is investigated whether lemmatization, spelling correction, and removal of stop words are suitable strategies for pre-processing within toxicity detection. The pre-trained DistilBERT model was fine-tuned using an English twitter dataset that had been pre-processed using a number of different techniques. The results indicate that none of the above-mentioned strategies have a positive impact on the model performance. Lastly, modern methods are applied to train a toxicity detection model adjusted to anonymous Mentimeter user text data. For this purpose, a balanced Mentimeter dataset with 3654 datapoints was created and annotated by the thesis author. The best-performing model of the pre-processing experiment was iteratively fine-tuned and evaluated with an increasing amount of Mentimeter data. Based on the results, it is concluded that state-of-the-art performance can be achieved even when using relatively few datapoints for fine-tuning. Namely, when using around 500 − 2500 training datapoints, F1-scores between 0.90 and 0.94 were obtained on a Mentimeter test set. These results show that it is possible to create a customized toxicity detection program, with high performance, using just a small dataset. / I och med sociala mediers stora framtåg har allvarliga problem såsom nätmobbning och spridning av hat online blivit allt mer vanliga. Mentimeter är ett interaktivt presentationsverktyg som gör det möjligt för presentations-publiken att svara på frågor genom att formulera egna fritextsvar. Eftersom Mentimeter ofta används i skolor så finns det ett behov av ett välfungerande program som identifierar och filtrerar ut kränkande text och svordomar. Den här uppsatsen fokuserar på ämnena textbehandling och robusthet gentemot dataskifte i en kontext av identifiering av kränkande språk för engelsk text. Först undersöks det huruvida lemmatisering, stavningskorrigering, samt avlägsnande av stoppord är lämpliga textbehandlingstekniker i kontexten av identifiering av kränkande språk. Den förtränade DistilBERT-modellen används genom att finjustera dess parameterar med ett engelskt Twitter-dataset som har textbehandlats med ett antal olika tekniker. Resultaten indikerar att ingen av de nämnda strategierna har en positiv inverkan på modellens prestanda. Därefter användes moderna metoder för att träna en modell som kan identifiera kränkande text anpassad efter anonym data från Mentimeter. Ett balancerat Mentimeter-dataset med 3654 datapunkter skapades och annoterades av uppsatsförfattaren. Därefter finjusterades och evaluerades den bäst presterande modellen från textbehandlingsexperimentet iterativt med en ökande mängd Mentimeter-data. Baserat på resultaten drogs slutsatsen att toppmodern prestanda kan åstadkommas genom att använda relativt få datapunkter för träning. Nämligen, när ungefär 500 − 2500 träningsdatapunkter används, så uppnåddes F1-värden mellan 0.90 och 0.94 på ett test-set av Mentimeter-datasetet. Resultaten visar att det är möjligt att skapa en högpresterande modell som identifierar kränkande text, genom att använda ett litet dataset.

Questo studio si concentra sull’analisi di un fenomeno assai comune nell’italiano e ben attestato da oltre un secolo in diverse altre lingue, antiche e moderne: le costruzioni prive di un verbo in forma finita nel loro nucleo sintattico principale, che evidentemente non sono state oggetto di una ellissi e che non sempre possono essere definite frasi. Dopo le analisi su questo fenomeno fatte da Mortara Garavelli (1971) sullo scritto letterario e da Cresti (1998) sul parlato colloquiale, in questa ricerca si vuole indagare la natura delle costruzioni senza verbo in una nuova varietà di italiano, ossia lo scritto informale e dialogico prodotto sul web, che sarà definito in questa ricerca come italiano digitato colloquiale (IDC). Pertanto, questo studio adotta un approccio corpus-based, ricercando le costruzioni senza verbo in una raccolta di testi di IDC realmente prodotti, ossia nel corpus COSMIANU (Corpus Of Social Media Italian Annotated with Nominal Utterances) (Comandini et al., 2018). Si è dunque deciso di individuare il fenomeno sulla base della definizione di enunciato nominale di Ferrari (2011; 2014), ma adottando due prospettive sintattiche ancora mai applicate in ambito italiano: la teoria sentenzialista di Merchant (2004; 2006; 2010) e quella non-sentenzialista di Barton & Progovac (2005), entrambe applicate in inglese a strutture ellittiche definite frammenti senza antecedente esplicito. Pertanto, si è deciso di definire le strutture senza verbo studiate come frammenti nominali, nell’ottica tanto di inquadrare un fenomeno che, nella nuova varietà di lingua studiata, assume forme diverse rispetto allo scritto letterario e al parlato colloquiale, quanto di unire simbolicamente due tradizioni di studio delle costruzioni senza verbo che non si sono mai incontrate, ossia quella italo-francese, risalente a Meillet (1906), e quella anglo-americana, risalente a Sweet (1900). Grazie all’analisi dei frammenti nominali nell’italiano digitato colloquiale in ottica non-sentenzialista, si so-no individuate undici classi di frammenti nominali, alcuni dei quali possono essere considerati delle frasi, poi-ché contengono o un rapporto predicativo tra due costituenti, o una Tense Phrase al proprio interno. Sul fronte dell’analisi sentenzialista, invece, si è ipotizzata l’esistenza di una nuova categoria di frammenti nominali, nei quali è stato eliso un elemento pro e un verbo essere. Grazie al contributo tanto della teoria sentenzialista, quanto di quella non-sentenzialista, è stato possibile notare come l’ IDC abbia come uno dei tanti tratti diagnostici proprio la presenza di frammenti nominali che ne incarnano le caratteristiche principali, ossia: a) l’estrema natura dialogica, che quindi spiega l’alta presenza di formule di saluto e di ringraziamento (es.: CIAO A TUTTE LE FANS; grazie 1000000000000) e di interiezioni (es.: bleah!); b) la forte aderenza al contesto comunicativo, con frammenti nominali che hanno come nodo iniziale un NP, un DP o un AP che fa direttamente riferimento a un elemento precedentemente reso rilevante nel contesto (es.: Bellissimoooooooooooo !!!!!!!!!!!!), oppure a un elemento immediatamente successivo, di cui si specifica la natura (es.: una domanda... perché é all'inverso?). Successivamente, si è testato come l’individuazione e l’analisi sintattica dei frammenti nominali possa aiutare a comprendere e a riconoscere meglio l’hate speech. Analizzando i frammenti nominali portatori d’odio nel corpus di tweet razzisti POP-HS-IT (Comandini & Patti, 2019), si è notato come l’ IDC d’odio presenti le medesime classi di frammenti nominali individuate in COSMIANU, ma in percentuali diverse, con una partico-lare rilevanza dei frammenti nominali che hanno come nodo iniziale un FocP (es.: FUORI QUESTE MERDE UMANE DALL'ITALIA). Inoltre, si è trovata una notevole presenza di frammenti nominali di classe FocP (es.: pezzi di merda loro e tutto l’islam) corrispondenti alle frasi esclamative studiate da Munaro (2006) (es.: Noioso, il tuo amico!), in cui l’elemento focalizzato a sinistra (pezzi di merda) è sempre una caratteristica intrinseca e non temporanea del soggetto (loro e tutto l’islam). Questa tipologia di frammenti nominali esclamativi e focalizzati veicola alcune delle caratteristiche più universali dell’hate speech, ossia l’espressione di un odio generalizzato e non dibattibile verso una categoria di persone vista come un gruppo monolitico. L’individuazione dei frammenti nominali più caratteristici dell’hate speech potrebbe aiutare i tool automatici ad annotare i testi d’odio in maniera più accurata.

This MA thesis analyzes how the protagonists and the narrative structure relates to “the Child” as a linguistic/cultural construction in Mare Kandre’s early authorship (1984-1991). A thematic of children portraits defines this period of Kandre’s writing, which consits of the following works: In a Different Country (1984), The Annunciation (1986), Bübin’s Kid (1987), The Burning Tree (1988) and Aliide, Aliide (1991). Linguistic components, such as the word “child”, the name “Kid” and a reconstruction of the concept of childhood, holds an actuate position in these narratives when they first appear in direct relation to the protagonists. This suggests how generally accepted categories and concepts consciously are at work in Kandre’s writings. Consequently, a category such as “child” can not unproblematically serve as a definition of any of the protagonists, irrespective of their presumptive age. Combining philosopher Judith Butler’s theory of name-calling with theories in the field of Childhood Studies as presented by Susan Honeyman, “the Child” in Kandre’s works is explored through the hypothesis of its function as a name. Close readings reveal how the protagonists, in the encounter with an explicitly cultural or narrative context, are constituted as subjects through the linguistic category “child”. The protagonists can be considered both as subordinated by the conventionally manifested notion of life development as a linear, measurable progress and as introduced to the means which make possible a critical response to that same notion. Since the main protagonists strongly refuse to be defined by the conventions supporting “the Child” as a category, a dialogue is created which gives the narrative its force. That dialogue corresponds to different aspects of the concept of childhood, which place “the Child” as a construction at the heart of Kandre’s narrative composition.

In an era when adversarial politics is condemned for either being archaic or right-wing extremism, proposing that incivility can be used to counter existing hegemonies, despite its potential to incite violence, is proposing an unorthodox project. By rejecting foundationalist approaches to the current incivility crisis, this study sees an opportunity for it to act as a populist rapture that defies simple binary categorisation and deconstructs incivility, at an ontological level, to reveal the deep meanings and concealed causes that contrast the grand narrative of hate speech. After an overview in chapter one, the study continues with a theoretical review of literature on incivility, guided by the works of radical democracy theorists who universalise what seems particular to Kenya. This review is followed by the description of Bakhtin’s concept of carnivalesque as utani, a joking relationship common in East Africa. For its theoretical perspective, the study is guided by Mouffe’s theory of agonistic democracy and a research method developed by transforming Laclau and Mouffe’s (1985) work in Hegemony and Socialist Strategy: Towards a Radical Democratic, into a method for Discourse Analysis. Various concepts from Laclau and Mouffe’s work are used to innovate an explanation of how political practices in social media, both linguistic and material texts, enhance incivility and the struggle to fix a regime’s preferred meaning. Guided by Laclau and Mouffe’s Discourse Analysis, the study describes how the government is using linguistic tools and physical technologies to repair the dislocation caused by incivility in social media in its attempts to re-create hegemonic practices. Without engaging in naïve reversal of the polarities between acceptable and unacceptable speech, and considering that at the ontological level politics is a friend—enemy relation, the study argues that incivility in social media is part of the return of politics in a post-political era, rather than simple unacceptable speech. While remaining aware of the dangers of extreme speech, but without reinforcing the anti-political rational consensus narrative, incivility is seen as having disruptive counterhegemonic potential, that is, if we consider the powerplay inherent in democracy. It means that binary opposition is blind to the way power produces, and is countered through unacceptable speech. / Communication Science / D. Litt. et Phil. (Communication Science)

In this thesis, I have studied the United Nations concluding observations on Sweden in relation to international human rights obligations. Specifically, in regards to the International Convention on the Elimination of All Forms of Racial Discrimination (ICERD) and the Convention on the Rights of the Child (CRC).  The Committee on the Rights of the Child have, in their Concluding Observations from 2023, expressed a deep concern about continued racism and discrimination of many children in Sweden. Meanwhile, the Committee on the Elimination of Racial Discrimination has recommended Sweden to take immediate measures for the purpose of eliminating all forms of racial discrimination.  In terms of critical analysis, the aim of this thesis has been to study how the Swedish state's international human rights obligations for the purpose of eliminating racial discrimination relates to the implementation of children's rights according to the principle of non-discrimination. According to my study, it is necessary for the Swedish state to promote and respect the human dignity inherent in every child, to be able to guarantee every child within the jurisdiction the rights under the CRC. A central example is the child's right to education in alignment with general human rights principles. Therefore, it is required for the state to prevent the normalization of racism in Sweden, in accordance with the international human rights obligations under the ICERD. This includes, for example, the prohibition of racist organizations in national law, for the purpose of eliminating racist hate speech, and the elimination of racial profiling by national law enforcement. Many children's living conditions and access to the rights under the CRC are affected by racism in several ways, through structural and everyday racism, when they themselves are exposed to racism and when people in their close surroundings experience racism. This has led me to the conclusion that active and immediate measures by the Swedish state, for the purpose of eliminating racial discrimination, are necessary for the possibility of guaranteeing the children who are victims of racism their rights in alignment with the CRC.

The topic of this thesis is the detoxification of language in social networks with a particular focus on style transfer techniques that combine deep learning and linguistic resources. In today’s digital landscape, social networks are rife with communication that can often be toxic, either intentionally or unintentionally. Given the pervasiveness of social media and the potential for toxic language to perpetuate negativity and polarization, this study addresses the problem of toxic language and its transformation into more neutral expressions. The importance of this issue is underscored by the need to promote non-toxic communication in the social networks that are an integral part of modern society. The complexity of natural language and the subtleties of what constitutes toxicity make this a challenging problem worthy of study. To address this problem, this research proposes two models, LexiconGST and MultiLexiconGST, developed based on the Delete&Generate framework. These models integrate linguistic resources into the detoxification system to guide deep learning techniques. Experimental results show that the proposed models perform commendably in the detoxification task compared to stateof-the-art methods. The integration of linguistic resources with deep learning techniques is confirmed to improve the performance of detoxification systems. Finally, this research has implications for social media platforms and online communities, which can now implement more effective moderation tools to promote non-toxic communication. It also opens lines of further research to generalize our proposed method to other text styles. / Ämnet för denna avhandling är avgiftning av språk i sociala nätverk med särskilt fokus på stilöverföringstekniker som kombinerar djupinlärning och språkliga resurser. I dagens digitala landskap är sociala nätverk fulla av kommunikation som ofta kan vara giftig, antingen avsiktligt eller oavsiktligt. Med tanke på hur utbredda sociala medier är och hur giftigt språk kan bidra till negativitet och polarisering, tar den här studien upp problemet med giftigt språk och hur det kan omvandlas till mer neutrala uttryck. Vikten av denna fråga understryks av behovet av att främja giftfri kommunikation i de sociala nätverk som är en integrerad del av det moderna samhället. Komplexiteten i naturligt språk och de subtila aspekterna av vad som utgör toxicitet gör detta till ett utmanande problem som är värt att studera. För att ta itu med detta problem föreslår denna forskning två modeller, LexiconGST och MultiLexiconGST, som utvecklats baserat på ramverket Delete&Generate. Dessa modeller integrerar språkliga resurser i avgiftningssystemet för att vägleda djupinlärningstekniker. Experimentella resultat visar att de föreslagna modellerna presterar lovvärt i avgiftningsuppgiften jämfört med toppmoderna metoder. Integrationen av språkliga resurser med djupinlärningstekniker bekräftas för att förbättra prestanda för avgiftningssystem. Slutligen har denna forskning konsekvenser för sociala medieplattformar och onlinegemenskaper, som nu kan implementera mer effektiva modereringsverktyg för att främja giftfri kommunikation. Det öppnar också för ytterligare forskning för att generalisera vår föreslagna metod till andra textstilar.

Freedom of expression is a fundamental human right which is important not only for the intellectual integrity of an individual but also for the healthy development of the whole society. The work represents both general arguments for freedom of expression and codification of freedom in two different legal cultures. In the United States the freedom of expression is protected by the First Amendment of the Constitution, which complements the rich jurisprudence of the Supreme Court of the United States. In Mexico, the right is regulated in Article 6 of the Constitution. The work shows how both countries reached the current legislation on the basis of examination of legal developments and cultural values. The values of the societies are presented on the basis of Hofstede's cultural dimensions. The core of the thesis is to compare the legal limits, which is made primarily on the basis of judicial practice of the Supreme Courts of both countries. Emphasis is placed on four main areas which are related to freedom of speech: (i) fighting words including hate speech, (ii) symbolic speech, (iii) obscenity, and (iv) defamation. In addition to the legal limits the work interprets also the most significant actual limits in both countries. Finally, the author examines the relationship between freedom of expression and...

Die hier vorgelegte kumulative Dissertation befasst sich mit der Frage des emotionalen Sprachgebrauchs als Teil der öffentlichen politischen Kommunikation in abendländischen Demokratien. Durch die Anwendung etablierter Erkenntnisse und Ansätze aus den Bereichen der affektiven Neurowissenschaften und der Neurolinguistik zu emotionalem Sprachgebrauch, wird versucht neue Perspektiven und Analysetechniken für die Sozial- und Politikwissenschaften herauszuarbeiten, die sich mit der rhetorischen Gestaltung und Funktion öffentlicher politischer Kommunikation befassen. Im Rahmen dieser Bemühungen habe ich einen sequenziellen Mixed-Methods-Ansatz entwickelt, eine neopragmatische Diskursanalyse (NPDA), die auf der bestehenden Methodologie der Pragmatischen Soziologie der Kritik (PSC) basiert. Diese Methode ermöglicht es, die Argumentationsstrategien und -muster von Akteuren abzuleiten und ihre Verwendung von hochgradig erregender emotionaler Sprache (via HAEWWörterbuch) nachzuzeichnen. In zwei separaten Fallstudien habe ich diesen neu entwickelten Ansatz angewendet, um (1) die US-Präsidentschaftswahlkämpfe 2016 von Trump und Clinton zu analysieren sowie (2) öffentliche Reden britischer Parteiführer:innen der Conservative und der Labour Party von 1900 bis 2019. Ich habe festgestellt, dass emotionale Sprache in der gesamten untersuchten (politischen) Kommunikation vorkommt und dass sie hauptsächlich eine Highlighter-Funktion in den Argumentationen der Akteure einnimmt. Politische Kommunikation, als eine spezifische Ausformung menschlicher Kommunikation, scheint immer ‚emotional‘ zu sein. / This cumulative dissertation addresses the issues of emotive language use as part of public political communication in occidental democracies. By applying established findings and approaches from the fields of affective neuroscience and neurolinguistics on emotive language use, the here presented dissertation intends to offer new perspectives and analytical techniques for the social and political sciences, concerned with understanding the rhetorical design and function of public political communication. As part of these efforts, I developed a sequential mixed methods approach, a neopragmatist discourse analysis (NPDA), which is based on the existing methodology associated with the Pragmatic Sociology of Critique (PSC). This method allows for the deduction of actors’ argumentative strategies and patterns and to detect their use of highly arousing emotive language (via HAEW dictionary). In two separate case studies I applied this newly developed approach, analysing (1) the 2016 U.S. presidential election campaigns of Trump and Clinton as well as (2) public speeches of British party leaders from the Conservative and Labour Party from 1900 – 2019. I found that emotive language appeared throughout all analysed (political) communication and that it mainly served a highlighting function within actors’ argumentations. Political communication, as a specific form of human communication, seems to always be ‘emotional’.