The rapid expansion of technology has posed challenges in moderating hateful content on social media platforms, particularly in non-text formats like TikTok videos. This study addresses a significant gap in scholarly research by examining the intersection of TikTok and journalistic content within the Ethiopian diaspora community. The research aims to investigate the prevalence and characteristics of hate speech in ethnic-based social media platforms operated by the Ethiopian diaspora, focusing on two prominent platforms: Ethio 360 Media and Gadaa Media Network (Gadaa MN), representing the Amhara and Oromo ethnic groups, respectively. Employing content analysis methodology, the study examines a hundred TikTok video clips from these platforms, drawing upon the theoretical framework of Framing Theory. This framework encompasses Frame Amplification, Frame Transformation, and Frame Contest, allowing for the identification of distinct patterns in hate speech language and content presentation between the two platforms. Key findings highlight several significant insights: (1) Both Ethio 360 Media and Gadaa MN exhibit similarities in hate speech, particularly in incitement of violence or threats. However, Ethio 360 Media employs uniquely demeaning and dehumanizing language, absent in Gadaa MN's journalistic content; (2) Analysis of hate speech types reveals differences in prevalence and focus within journalistic content, with both platforms engaging in politics-based hate speech but disparities emerging in ethnicity-based hate speech; (3) Examination of targeted entities unveils differing framing strategies, with both platforms targeting the Ethiopian government but diverging in their approaches toward other entities such as the Oromo ethnic group and educational institution; (4) Analysis of news coverage genres indicates variations in content presentation, with Ethio 360 Media relying more on commentary while Gadaa MN focuses predominantly on straight newscasts; (5) Exploration of news coverage areas reveals distinct content focuses between the two platforms, with Ethio 360 Media covering Ethiopia as a whole and various regions within the country, while Gadaa MN primarily focuses on the Oromia Region, occasionally covering foreign countries or regions.

Univ., Diss.--Würzburg, 2001. / Literaturverz. S. 363 - 380.

Given the explosion in the size of social media, the amount of hate speech is also growing. To efficiently combat this issue we need reliable and scalable machine learning models. Current solutions rely on crowdsourced datasets that are limited in size, or using training data from self-identified hateful communities, that lacks specificity. In this thesis we introduce a novel semi-supervised modelling strategy. It is first trained on the freely available data from the hateful communities and then fine-tuned to classify hateful tweets from crowdsourced annotated datasets. We show that our model reach state of the art performance with minimal hyper-parameter tuning.

This study was conducted to identify and analyse Mail and Guardian Online moderation outputs which contradicted the platform‟s own stated policy on hate speech and other forms of problematic speech. The moderation outputs considered were a battery of readers‟ comments that were posted in response to Thought Leader blogs on xenophobia published between May and June, 2008. This was the same period a series of xenophobic attacks was taking place in some parts of South Africa, leaving an estimated 62 people dead, more than 30,000 displaced, and countless victims injured and robbed of their property. The attacks were a catalytic moment that enabled a whole range of discursive positions to be articulated, defended, contested and given form in the media. They also made visible the potential tensions between free speech on the one hand, and hate and other problematic speech on the other. Using qualitative methods of thematic content analysis, document review, individual interviews, and an eclectic approach of framing analysis and rhetorical argumentation, this study found instances of divergence between the M&G policy and practice on User Generated Content. It found that some moderator-approved content advocated hate, hatred, hostility, incitement to violence and/or harm, and unfair discrimination against foreign residents, contrary to the M&G policy which is informed by the constitutional provisions in both section 16 of the Bill of Rights and section 10 of the Equality Act. Based on examples in the readers‟ comments of how „the foreigner‟ was made to signify unemployment, poverty, disease, unfair competition, and all manner of deprivation, and bearing in mind how such individuals have also become a site for the violent convergence of different unresolved tensions in the country, the study‟s findings argue that the M&G – a progressive paper dealing with a potentially xenophobic readership (at least a portion of it) – should have implemented its policy on acceptable speech more effectively. The study also argues that the unjustifiable reference to foreigners as makwerekwere, illegals, illegal aliens, parasites, invaders and border jumpers, among other terms, assigned them a diminished place – that of unwanted foreigner – thereby reproducing the order of discourse that utilises nationality as a space for the expurgation of the „other‟. The study argues that the use of bogus (inflated) immigration statistics and repeated reference to the foreigners‟ supposedly parasitic relationship to the country‟s resources also unfairly constructed them as the „threatening other‟ and potentially justified action against them.

This paper concerns the freedom of speech in our democracy and the principles underlying the restrictions. The study is aimed at using a descriptive theory to provide a useful basis for analyzing a justification of the role behind freedom of speech and its limitations. The analysis is made on the basis of the Swedish parliament debate in the form of motions and the non-governmental organization Swedish Helsinki Committee's report. The study has a focus on the law of hate speech and will be done with an idea centred analysis. On the basis of democratic theory, the central role of freedom of speech was clarified on the basis of a clear need for a reliable communication. An absence of this process could disrupt the democratic order. The theoretical function of restrictions is to maintain that reliable communication process where, for example, special circumstances, threats of violence and certain types of information were seen as examples of situations where the process could be in danger. The empirical result was largely in line with the theoretical conclusions. The ability to argue and keep open debates was seen as important building blocks in a democratic society and therefore implies the benefits of a reliable communication process. Justification for the restrictions was made with regard to harmful information, threatening circumstances and xenophobic organizations, who were examples of counter-productive expressions to the democratic order. However, the study showed that parliamentary motions and the Swedish Helsinki Committee often focused their justifications on different fundamental views which partly could be explained by the concepts of positive and negative freedom and the perception of minority rights. The findings indicated that the issue of restrictions on freedom of speech is extremely complex because of the varied opinions in the drafting of the law of hate speech. The dilemma of freedom of expression can therefore be seen as relevant up to this day.

Cette thèse examine, depuis une perspective durkheimienne, le conflit politique en Allemagne sur le multiculturalisme, l’immigration, et l’identité nationale. L’analyse se fait dans la période entre le début de la crise des réfugiés en 2015 et l’élection fédérale allemande de septembre 2017. J’identifie quatre idéaux moraux, soit des visions idéales de la communauté allemande qui motivent les acteurs politique : l’idéal des Autonomen qui rejettent tout forme de pouvoir et de domination, l’idéal du Verfassungspatriotismus (le patriotisme constitutionnel) de la SPD (et une partie de la CDU), l’idéal de la nation (ethno)culturelle de la plupart de l’AfD (et la CSU et la WerteUnion), et l’idéal de la nation biologique de l’extrême droite. Au cœur de chaque idéal est un objet sacré qui sert d’autorité morale qui légitimant des prescriptions morales et qui amène à une série de vérités morales et de jugements moraux, la totalité duquel Émile Durkheim identifie comme un fait moral. Pour les Autonomen et les adhérents du Verfassungspatriotismus, l’objet sacré est l’individu conçu à travers le concept de la dignité humaine. Pour les autres, l’objet sacré est la nation allemande, conçue en termes (ethno)culturels ou en termes biologiques. Cette thèse argumente que ces idéaux moraux sont intrinsèquement profanatoires, dans le sens que les prescriptions morales d’un objet sacré (la dignité humaine) violent directement l’objet sacré de l’autre (la nation), et vice-versa. Ces idéaux sont tous en concurrence pour le pouvoir et l’influence, avec comme but d’avoir accès au pouvoir étatique allemand. Le résultat est un conflit politique qui traduit essentiellement un conflit moral. Ces conflits ont lieu dans le domaine légal, au sein des partis politique, et à travers la violence politique. Ces conflits touchent un nombre de sujets clés comme la liberté d’expression, le multiculturalisme, et l’extrémisme politique. La présente thèse cherche à comprendre ces conflits à travers le prisme du concept durkheimien du fait moral, et développe une sociologie du conflit moral durkheimien. Cette thèse s’inspire également de la théorie de conflit de Randall Collins, qui s’inspire elle aussi de l’œuvre de Durkheim. / This dissertation examines, from a Durkheimian perspective, political conflict in Germany around the issues of multiculturalism, immigration, and national identity within the context of the Refugee Crisis beginning in 2015 and ending roughly with the German Federal Election in September 2017. It identifies four moral ideals, or ideal visions of the German community, that motivated political actors during this period: the Autonomen ideal that rejects all forms of power and domination; the ideal of Verfassungspatriotismus (Constitutional Patriotism) of the SPD (and parts of the CDU); the ideal of the cultural or ethnocultural nation of much of the AfD (and the CSU and WerteUnion); and the ideal of the biological nation on the far-right. At the heart of each moral ideal is a sacred object that serves as a moral authority that legitimates certain moral prescriptions, and leads to a set of moral truths and moral judgments, the totality of which Émile Durkheim identifies as a moral fact. For the Autonomen and adherents of Verfassungspatriotismus the sacred object is the individual understood through the concept of human dignity. For others the sacred object is the German nation, understood either in an (ethno)cultural sense or a biological sense. As the dissertation argues, these different moral ideals are inherently profanatory to each other, such that the moral prescriptions inspired by one sacred object (human dignity) directly violate the sacred object of the other (the nation), and vice-versa. These ideals all compete with each other for power and influence within the German political sphere as a means to gain access to (or to dismantle) state power. The result is political conflict that takes place essentially within a moral framework. These conflicts occur in the legal domain, in battles over party leadership and membership, and through political violence; they touch on a number of key issues such as free speech, multiculturalism, and political extremism. This dissertation seeks to understand these conflicts through the prism of Durkheim’s concept of the moral fact and to develop a Durkheimian sociology of moral conflict. In this analysis, the dissertation draws on Randall Collins’ conflict theory, which Durkheim’s work also largely inspires.

El objetivo del siguiente trabajo es analizar cómo se manifiestan y regulan los discursos de odio en redes sociales relacionados al clasismo étnico en el Perú. En ese sentido, la pregunta de investigación es la siguiente: ¿Cómo se manifiestan y regulan los discursos de odio en redes sociales relacionados al clasismo étnico en el Perú? Para poder responder esta pregunta se han explicado diversos términos dentro del marco teórico tales como: Discursos de odio, redes sociales, clasismo étnico en el Perú, libertad de expresión, etc. Cada uno guarda relación con el otro para poder comprender de forma general el tema y cómo existen variantes que determinan la expansión y poca regulación del discurso de odio con contenido racista y clasista. Para ello, se ha realizado un diseño metodológico basado en el análisis de contenido con un paradigma de constructivismo social, y se han realizado instrumentos como una matriz de análisis de contenido para clasificar los comentarios de los contenidos analizados y una guía de entrevista para poder conocer la opinión de los usuarios que pertenecen a esta comunidad sobre este tipo de contenidos y la gestión que tiene Facebook para censurarlos o permitir que sigan circulando indiscriminadamente. / The objective of the following work is to analyze how hate speech is manifested and regulated in social networks related to ethnic classism in Peru. In this sense, the research question is the following: ¿How is hate speech in social networks related to ethnic classism manifested and regulated in Peru? In order to answer this question, various terms have been explained within the theoretical framework such as: hate speech, social networks, ethnic classism in Peru, freedom of expression, etc. Each one is related to the other in order to understand the subject in a general way and how there are variants that determine the expansion and little regulation of hate speech with racist and class content. For this, a methodological design based on content analysis with a paradigm of social constructivism has been carried out, and instruments have been made such as a content analysis matrix to classify the comments of the analyzed content and an interview guide to be able to know the opinion of the users who belong to this community about this type of content and the management that Facebook has to censor them or allow them to continue circulating indiscriminately. / Trabajo de investigación

Dans le contexte actuel de la polarisation sur les réseaux sociaux, ce mémoire de maîtrise porte sur la négociation à plusieurs niveaux des normes sociales relatives au discours haineux, sur le forum de Reddit. La recherche s'articule autour de la régulation du contenu par les modérateurs et les utilisateurs. Le mémoire commence par présenter le forum Reddit, son histoire et son fonctionnement avant de développer une revue de littérature sur le discours de haine et le misogynie, notamment l'absence de consensus sur sa définition ainsi que les défis juridiques et opérationnels de sa régulation. Ensuite, pour comprendre comment les utilisateurs et les modérateurs de Reddit régulent les discours de haine misogyne, le cadre conceptuel développe le concept de normes sociales, en particulier les normes sociales en ligne. Parmi les autres concepts mobilisés figurent les chambres d'écho et la théorie de l'étiquetage. Une approche qualitative a été adoptée pour faire une analyse schématique d’un corpus de données empiriques publiquement disponibles sur Reddit qui a été sélectionné pour son lien avec le discours haineux misogyne. Cinq communautés (« subreddits ») ont été sélectionnées pour explorer la manière dont les discours misogynes sont régulés. Les résultats présentent une typologie des actions disponibles pour les utilisateurs et les modérateurs, ainsi que les règles de Reddit et de chaque communauté sur lesquelles ils et elles s'appuient pour réguler le discours haineux misogyne. La typologie d'actions est représentée au sein de schémas répétitifs qui mettent en valeur la façon dont la régulation se produit sur les communautés. Plusieurs exemples illustratifs sont offerts de ces négociations réglementaires. La recherche conclue que l'étiquetage par les actions répétitives pourrait potentiellement contribuer à la négociation des normes sociales entre les utilisateurs et les modérateurs et ainsi avoir des effets sur la façon dont les échanges se produisent sur les fils de discussions, au sein du forum. Cette négociation des normes ferait donc un lien direct quant à la fine ligne qui subsiste entre la liberté d'expression et le discours haineux, et recadrer ce qui en constitue ou non, pour comprendre les types de discours acceptables sur cette plateforme. / In the current context of polarization on social networks, this Master's thesis examines the multi-level negotiation of social norms around hate speech on the Reddit forum. The research focuses on the regulation of content by moderators and users. The thesis begins by introducing the Reddit forum, its history and operations before developing a literature review on hate speech and misogyny, including a lack of consensus on its definition and the legal and operational challenges of regulating it. Then, to better understand how Reddit users and moderators regulate misogynous hate speech, the conceptual framework develops the concept of social norms, specifically online social norms. Other concept of echo chambers and labeling theory are also mobilized. The research adopts a qualitative approach to conduct schematic analysis of a body of publicly available empirical data on Reddit that was selected for its connection to misogynistic hate speech. Five communities (“subreddits”) were selected to explore how misogynistic speech is regulated. In the findings, I present a typology of actions available to users and moderators, as well as the rules of Reddit and each community that they rely on to regulate misogynistic hate speech. The typology of actions is represented within repetitive patterns that highlight how regulation occurs on communities. Several illustrative examples are offered of these regulatory negotiations. The research concludes that labeling through repetitive actions could potentially contribute to the negotiation of social norms between users and moderators and thus have effects on the way exchanges occur on threads within the forum. This negotiation of norms would thus directly relate to the fine line that remains between free speech and hate speech, thus reframing what does and does not constitute hate speech, in order to understand the types of speech that are acceptable on this platform.

This thesis aims to explore the representation of transgender individuals in online Czech discourses. The study utilizes corpus linguistics to analyze news media and Facebook in order to understand the attitudes towards transgender individuals in these discourses. Previous research on gender in the Czech language and other languages is considered in order to contextualize the findings of this study. The research reflects on the Czech Press Act and Audio-visual Act, which provide guidelines for objectivity and balanced reporting in the media. Using data from Czech National Corpus the study aims to shed light on the representation of transgender individuals in online Czech discourses and provide exposure of potential hateful speech towards trans people or imbalance between the discourses. / Denna studie avser att undersöka representationer av transpersoner i webbaserade diskurser i den tjeckiska språkmiljön. I uppsatsen används korpuslingvistik för analys av nyhetstexter och Facebook med syftet att förstå vilka attityder mot transpersoner som förekommer i dessa diskurser. Tidigare lingvistisk forskning med tjeckiska korpusbaserade diskursanalyser i fokus tas i beräkning och kontextualiseras i denna studie. Tjeckisk lagstiftning kring media och objektivitetskrav reflekteras i forskningen. Denna uppsats ämnar exponera och analysera representationer av transpersoner i olika diskurser samt reflektera över huruvida kraven som den tjeckiska lagstiftningen ställer speglas i dessa onlinemiljöer.

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.