AI and Linguistic Synergies for Advanced Detection of Online Extremism

Prof Encarna Hidalgo-Tenorio

Professor in English Linguistics, University of Granada and CDH Visiting Fellow

Encarna Hidalgo-Tenorio’s research career, spanning 30 years as a Research Fellow, University Lecturer, and Full Professor, focuses on language use in literature (particularly Irish theatre), politics, and media. Her work has explored the intersection of language with gender, critical linguistics, discourse analysis, corpus linguistics, and systemic-functional linguistics. Notably, her research examines concepts such as transitivity, modality, and evaluation to understand the strategies speakers use to represent their realities and social actors, thereby shaping their own and others’ identities. She has critiqued and sometimes refuted established theories through her publications. For instance, she has questioned the validity of the notion ‘genderlect’ based on real observations; refined Halliday’s taxonomy of processes and roles with more detailed classifications; and redefined categories such as transitivity, evidentiality, and modality, enhancing the interpersonal metafunction.

The purpose of this presentation is to outline the motivations, foundations, and methodological principles underlying the efforts of linguists and computer scientists in developing Nutcracker, a semi-supervised algorithm designed to detect extremist profiles and misinformation on social media. Nutcracker, accessible at https://nutcracker.ugr.es, is the result of a transdisciplinary collaboration initiated through various research projects led by the University of Granada since 2017.

This algorithm, based on what we term "deep relations" (Francisco & Castro, 2020), can perform the following tasks: (1) distinguish extremist profiles from non-extremist ones; (2) identify potential misinformation spreaders from those uninvolved in disseminating misleading content; and (3) map networks of users through linguistic patterns that reflect shared attitudes and sentiments (Francisco & Castro, 2020; Francisco, Benítez-Castro, Hidalgo-Tenorio & Castro, 2022). Nutcracker is supported by diverse ontologies meticulously identified by the team’s discourse analysts (Dhiab-Hassan, Benítez-Castro & Hidalgo-Tenorio, 2018). One of the most complex to articulate is derived from Martin & White’s Appraisal Theory (2005), incorporating a revision inspired by the taxonomy proposed in Benítez-Castro & Hidalgo-Tenorio (2019).

By integrating concepts from linguistic analysis with advanced computational techniques, this algorithm provides an example of how collaborative efforts across diverse fields can effectively address complex challenges at the intersection of technology and social issues, such as combating radicalisation or hate towards migrants.

Radicalisation is influenced by various socio-psychological factors, with the Internet’s anonymity and instant connectivity making it an ideal environment. This is one reason why jihadist groups have harnessed social media as a powerful tool for spreading propaganda. Their messaging is filled with persuasive pathos, setting in motion a process of mediated disorder (Aggarwal, 2019) intended to convince vulnerable individuals of the necessity of violence. Our algorithm has shown considerable effectiveness in detecting extremist content on Twitter. On the other hand, in a climate increasingly shaped by populism (see Hidalgo-Tenorio, Benítez-Castro, and De Cesare, 2019; Norris and Inglehart, 2019; Diehl and Bargetz, 2024) and polarisation (see Rekker and Harteveld, 2022; Bernaerts, Blanckaert, and Caluwaerts, 2023; Rojo-Martínez and Crespo-Martínez, 2023), misinformation is a pressing concern for politicians, the media, and society at large (see Espaliu-Berdud, 2023; Jungherr and Rauchfleisch, 2024; Nieminen, 2024). Misinformation has a powerful impact on public attitudes, perceptions, and behaviours and has even triggered political crises in the West, along with a pervasive sense of global instability (see Lemke and Habegger, 2022; Martinico and Monti, 2023; Vasist and Krishnan, 2023). We believe that Nutcracker offers a means to explore the complex interplay between migration and misinformation, shedding light on the multifaceted dynamics that appear to shape public opinion and deepen distrust towards “the Other” (see Díaz, Mejías, and Silva, 2023; Savatic, Thiollet, Mesnard, Senne, and Jaulin, 2024).

If you have specific accessibility needs for this event please get in touch. We will do our best to accommodate any requests.

Host or Publisher Cambridge Digital Humanities