Online Library

With the ever-increasing cases of hate spread on social media platforms, it is critical to design abuse detection mechanisms to pro-actively avoid and control such incidents. While there exist methods for hate speech detection, they stereotype words and hence suffer from inherently biased training. Bias removal has been traditionally studied for structured datasets, but we aim at bias mitigation from unstructured text data. In this paper, we make two important contributions. First, we systematically design methods to quantify the bias for any model and propose algorithms for identifying the set of words which the model stereotypes. Second, we propose novel methods leveraging knowledge-based generalizations for bias-free learning. Knowledge-based generalization provides an effective way to encode knowledge because the abstraction they provide not only generalizes content but also facilitates retraction of information from the hate speech detection classifier, thereby reducing the imbalance. We experiment with multiple knowledge generalization policies and analyze their effect on general performance and in mitigating bias. Our experiments with two real-world datasets, a Wikipedia Talk Pages dataset (WikiDetox) of size ~ 96k and a Twitter dataset of size ~ 24k, show that the use of knowledge-based generalizations results in better performance by forcing the classifier to learn from generalized content. Our methods utilize existing knowledge-bases and can easily be extended to other tasks.

Hate speech may take different forms in online social media. Most of the investigations in the literature are focused on detecting abusive language in discussions about ethnicity, religion, gender identity and sexual orientation. In this paper, we address the problem of automatic detection and categorization of misogynous language in online social media. The main contribution of this paper is two-fold: (1) a corpus of misogynous tweets, labelled from different perspective and (2) an exploratory investigations on NLP features and ML models for detecting and classifying misogynistic language.

In this study, we examine key psychological dimensions in the manifestos authored by the perpetrators of the Christchurch and Utøya massacres, the right-wing extremists Brenton Tarrant and Anders Breivik, and the ISIS propaganda magazine Rumiyah. All texts were authored and disseminated virtually with the purpose of attracting or consolidating support, and justifying violent, discriminatory actions. While right-wing supremacist and extremist Islamist discourses are ostensibly ideologically opposed, previous research has posited the existence of ideational and emotive commonalities. We approach this from a corpus-linguistic perspective, and employ the software LIWC2015 and Wmatrix to explore the dominant psychological dimensions, semantic categories and keywords in these texts. We identify elements that contribute to the construction of a narrative of hate, peril and urgency, and discuss differences in the imagery used to construct these meanings and to appeal to different audiences. Whilst our analysis supports the existence of commonalities in ideological content and discursive strategies, our results identify differences in the target of hate in right-wing supremacist discourse and we differentiate between primarily Islamophobic and racist motives. Finally, we also discuss the limitations inherent in employing these software tools to analyse discourse in the Web 2.0 era.

The number of lone-actor attacks committed by far-right extremists have surged in recent years, most notably in the West where mass-casualty attacks have occurred, including the United States, New Zealand, the United Kingdom, and Germany. The fatal attack in February 2020 in Hanau,
Germany, revealed the perpetrator’s influences to be a combination of traditional far-right, race-based, and anti-immigration narratives, alongside several more obscure conspiracy theories. This case demonstrates the need for further research into the intersection of these ideas and the online ecosystems in which they thrive, where notions such as the “Great Replacement” theory, aspects of which were echoed in the Hanau attacker’s own manifesto, are heavily propagated. It is this overarching idea that connects seemingly disparate attacks in a global network of ideologically analogous acts of terror.

This article seeks to propose a valuable frame for understanding which processes, either cognitive or practical, take part in the making of a terrorist act within the frame of the onlife region. A term recently coined, onlife, refers to the interacting/indistinguishable domain of online and offline world: this is the environment where today’s global terrorism/extremism flourishes. This region of complexity and the phenomena that entails can valuably explored through the General System Theory (GST). In the first part of the paper how global terrorism is changing, moving into cyber terrorism, at the light of the GST, according to recent theoretical achievements in the field is described in detail. The second part will explore a case study of onlife terrorism perpetrated in New Zealand, in order to focus the passage from the ‘possible’ idea of making a massacre to its effective performance, from onlife radicalization to the onlife event (a livestreamed attack filmed by the author) – through a deliberate use of the internet. At this point, the author will be able to come back to the question posed before, sketching some insights that might be valuable for cyber terrorism as well as for GST, especially for what concerns systemic processes within the moral domain.

In this paper, I analyze how the so called ‘alt-right’ is using Twitter to mainstream its politics. Understanding alt-right mainstreaming is important because the movement has embraced a Gramscian view of politics that believes cultural change (e.g. normalizing unpopular ideas) must precede institutional change (e.g. fielding candidates for office). To conduct my analysis I created a sample of 1,000 tweets from six ‘alt-right’ leaders. I then asked two questions about the tweets in this sample: what topics did alt-right leaders talk about most frequently, and how did they talk about them. My findings suggest that the ‘alt-right’s’ mainstreaming efforts on Twitter involve a mix of techniques. In terms of race, the alt-right is trying to normalize white identity politics. However, the movement is turning away from blatant misogyny on Twitter, instead strategically mimicking conservative tropes about women needing male protection. Finally, the alt-right references Europe more frequently than the US, suggesting that Europe is a geographic anchor in alt-right discourse.