Search results for: Marinol K. Hate

Authors: Julia Ive, Lucia Specia

Abstract:

Understanding and detecting toxicity is an important problem to support safer human interactions online. Our work focuses on the important problem of contextual toxicity detection, where automated classifiers are tasked with determining whether a short textual segment (usually a sentence) is toxic within its conversational context. We use “toxicity” as an umbrella term to denote a number of variants commonly named in the literature, including hate, abuse, offence, among others. Detecting toxicity in context is a non-trivial problem and has been addressed by very few previous studies. These previous studies have analysed the influence of conversational context in human perception of toxicity in controlled experiments and concluded that humans rarely change their judgements in the presence of context. They have also evaluated contextual detection models based on state-of-the-art Deep Learning and Natural Language Processing (NLP) techniques. Counterintuitively, they reached the general conclusion that computational models tend to suffer performance degradation in the presence of context. We challenge these empirical observations by devising better contextual predictive models that also rely on NLP data augmentation techniques to create larger and better data. In our study, we start by further analysing the human perception of toxicity in conversational data (i.e., tweets), in the absence versus presence of context, in this case, previous tweets in the same conversational thread. We observed that the conclusions of previous work on human perception are mainly due to data issues: The contextual data available does not provide sufficient evidence that context is indeed important (even for humans). The data problem is common in current toxicity datasets: cases labelled as toxic are either obviously toxic (i.e., overt toxicity with swear, racist, etc. words), and thus context does is not needed for a decision, or are ambiguous, vague or unclear even in the presence of context; in addition, the data contains labeling inconsistencies. To address this problem, we propose to automatically generate contextual samples where toxicity is not obvious (i.e., covert cases) without context or where different contexts can lead to different toxicity judgements for the same tweet. We generate toxic and non-toxic utterances conditioned on the context or on target tweets using a range of techniques for controlled text generation(e.g., Generative Adversarial Networks and steering techniques). On the contextual detection models, we posit that their poor performance is due to limitations on both of the data they are trained on (same problems stated above) and the architectures they use, which are not able to leverage context in effective ways. To improve on that, we propose text classification architectures that take the hierarchy of conversational utterances into account. In experiments benchmarking ours against previous models on existing and automatically generated data, we show that both data and architectural choices are very important. Our model achieves substantial performance improvements as compared to the baselines that are non-contextual or contextual but agnostic of the conversation structure.

Keywords: contextual toxicity detection, data augmentation, hierarchical text classification models, natural language processing

Procedia PDF Downloads 171

Authors: Mirjana Dokmanovic

Abstract:

The main topic of this paper is the analysis of the presence of gender stereotypes in the media content in the Republic of Serbia with respect to the state commitments to eliminate discrimination against women. The research methodology included the analysis of the media content of six daily newspapers and two magazines on the date of 28 December 2015 and the analysis of the reality TV show programs in 2015 from gender perspective. The methods of the research has also included a desk research and a qualitative analysis of the available data, statistics, policy papers, studies, and reports produced by the government, the Ministry of Culture and Information, the Regulatory Body for Electronic Media, the Press Council, the associations of media professionals, the independent human rights bodies, and civil society organizations (CSOs). As a State Signatory to the Convention on the Elimination of All Forms of Discrimination against Women, the Republic of Serbia has adopted numerous measures in this field, including the Law on Equality between Sexes and the national gender equality strategies. Special attention has been paid to eliminating gender stereotypes and prejudices in the media content and portraying of women. This practice has been forbidden by the Law on Electronic Media, the Law on Public Information and Media, the Law on Public Service Broadcasting and the Bylaw on the Protection of Human Rights in the Provision of Media Services. Despite these commitments, there has not been achieved progress regarding eliminating gender stereotypes in the media content. The research indicates that the media perpetuate traditional gender roles and patriarchal patterns. Female politicians, entrepreneurs, academics, scientists, and engineers have been very rarely portrayed in the media. On the other side, women are in their focus as celebrities, singers, and actresses. Women are underrepresented in the pages related to politics and economy, while they are mostly present in the cover stories related to show-business, health care, family and household matters. Women are three times more than men identified on the basis of their family status, as mothers, wives, daughters, etc. Hate speech, misogyny, and violence against women are often present in the reality TV shows. The abuse of women and their bodies in advertising is still widely present. The cases of domestic violence are still presented with sensationalism, although there has been achieved progress in portraying victims of domestic violence with respect and dignity. The issues related to gender equality and the position of the vulnerable groups of women, such as Roma women or rural women, are not visible in the media. This research, as well as warnings of women’s CSOs and independent human rights bodies, indicates the necessity to implement legal and policy measures in this field consistently and with due diligence. The aim of the paper is to contribute eliminating gender stereotypes in the media content and advancing gender equality.

Keywords: discrimination against women, gender roles, gender stereotypes, media, misogyny, portraying women in the media, prejudices against women, Republic of Serbia

Procedia PDF Downloads 206

Authors: Georgios Samaras

Abstract:

New media such as Facebook, YouTube and Twitter introduced the world to a new era of instant communication. An era where online interactions could replace a lot of offline actions. Technology can create a mediated environment in which participants can communicate (one-to-one, one-to-many, and many-to-many) both synchronously and asynchronously and participate in reciprocal message exchanges. Currently, social networks are attracting similar academic attention to that of the internet after its mainstream implementation into public life. Websites and platforms are seen as the forefront of a new political change. There is a significant backdrop of previous methodologies employed to research the effects of social networks. New approaches are being developed to be able to adapt to the growth of social networks and the invention of new platforms. Golden Dawn was the first openly neo-Nazi party post World War II to win seats in the parliament of a European country. Its racist rhetoric and violent tactics on social networks were rewarded by their supporters, who in the face of Golden Dawn’s leaders saw a ‘new dawn’ in Greek politics. Mainstream media banned its leaders and members of the party indefinitely after Ilias Kasidiaris attacked Liana Kanelli, a member of the Greek Communist Party, on live television. This media ban was seen as a treasonous move by a significant percentage of voters, who believed that the system was desperately trying to censor Golden Dawn to favor mainstream parties. The shocking attack on live television received international coverage and while European countries were condemning this newly emerged neo-Nazi rhetoric, almost 7 percent of the Greek population rewarded Golden Dawn with 18 seats in the Greek parliament. Many seem to think that Golden Dawn mobilised its voters online and this approach played a significant role in spreading their message and appealing to wider audiences. No strict online censorship existed back in 2012 and although Golden Dawn was openly used neo-Nazi symbolism, it was allowed to use social networks without serious restrictions until 2017. This paper used qualitative methods to investigate Golden Dawn’s rise in social networks from 2012 to 2019. The focus of the content analysis was set on three social networking platforms: Facebook, Twitter and YouTube, while the existence of Golden Dawn’s website, which was used as a news sharing hub, was also taken into account. The content analysis included text and visual analyses that sampled content from their social networking pages to translate their political messaging through an ideological lens focused on extreme-right populism. The absence of hate speech regulations on social network platforms in 2012 allowed the free expression of those heavily ultranationalist and populist views, as they were employed by Golden Dawn in the Greek political scene. On YouTube, Facebook and Twitter, the influence of their rhetoric was particularly strong. Official channels and MPs profiles were investigated to explore the messaging in-depth and understand its ideological elements.

Keywords: populism, far-right, social media, Greece, golden dawn

Procedia PDF Downloads 149

Authors: Vaiddehi Bansal, Laura Hinson, Mayumi Rezwan, Erin Leasure, Mithila Iyer, Connor Roth, Poulomi Pal, Kareem Kysia

Abstract:

As digital usage becomes increasingly ubiquitous worldwide, technology-facilitated gender-based violence (GBV) has garnered increasing attention in the recent years, especially during the COVID-19 pandemic. This form of violence is defined as “action by one or more people that harms others based on their sexual or gender identity or by enforcing harmful gender norms. This action is carried out using the internet and/or mobile technology that harms others based on their sexual or gender identity or by enforcing harmful gender norms”.Common forms of technology-facilitated GBV include cyberstalking, cyberbullying, sexual harassment, image-based abuse, doxing, hacking, gendertrolling, hate speech, and impersonation. Most literature on this pervasive yet complex issue has emerged from high-income countries, and few studies comprehensively summarize its prevalence, manifestations, and implications. This rigorous scoping review examines the evidence base of this phenomenon in low and middle-income countries across Asia, summarizing trends and gaps to inform actionable recommendations. The research team developed search terms to conduct a comprehensive search of peer-reviewed and grey literature. Query results were eligible for inclusion if they were published in English between 2006-2021 and with an explicit emphasis on technology-facilitated violence, gender, and the countries of interest in the Asia region. Title, abstracts, and full-texts were independently screened by two reviewers based on inclusion criteria, and data was extracted through deductive coding. Of 2,042 articles screened, 97 met inclusion criteria. The review revealed a gap in the evidence-base in Central Asia and the Pacific Islands. Findings across South and Southeast Asia indicate that technology-facilitated GBV comprises various forms of abuse, violence, and harassment that are largely shaped by country-specific societal norms and technological landscapes. The literature confirms that women, girls, and sexual minorities, especially those with intersecting marginalized identities, are often more vulnerable to experiencing online violence. Cultural norms and patriarchal structures tend to stigmatize survivors, limiting their ability to seek social and legal support. Survivors are also less likely to report their experience due to barriers such as lack of awareness of reporting mechanisms, the perception that digital platforms will not address their complaints, and cumbersome reporting systems. The COVID-19 pandemic has further exacerbated perpetration and strained support mechanisms. Prevalence varies by the form of violence but is difficult to estimate accurately due to underreporting and disjointed, outdated, or non-existent legal definitions. Addressing technology-facilitated GBV in Asia requires collective action from multiple actors, including government authorities, technology companies, digital and feminist movements, NGOs, and researchers.

Keywords: gender-based violence, technology, online sexual harassment, image-based abuse

Procedia PDF Downloads 134