Show simple item record

dc.rights.licenseCC-BY-NC-ND
dc.contributor.advisorVerhaar, P.
dc.contributor.authorEsser, A.C.
dc.date.accessioned2021-08-25T18:00:13Z
dc.date.available2021-08-25T18:00:13Z
dc.date.issued2021
dc.identifier.urihttps://studenttheses.uu.nl/handle/20.500.12932/41188
dc.description.abstractIn light of the unprecedented 2021 Capitol riots and the 2020 U.S. presidential election, there is no doubt that social media technologies like Parler facilitated online radicalization of individuals towards violence. As a result, the content on Parler, its lax positioning on content regulation, and heavily one-sided political leanings rendered it an attractive environment for empirically studying normative behaviors of a social network that is predominantly right-wing extremist. This thesis project considers characterizing the use of language by sub-communities detected on Parler composed of a conglomeration of conservatives, conspiracy theorists, and far-right extremists under a common goal to spread radical propaganda and disinformation on a mainstream social network. Through implementation of HDBSCAN clustering of Parler users into sub-communities based on the similarity of their hashtag patterns, it was possible to identify virtual social groups that subscribed to a wide assortment of far-right beliefs in varying degrees. Drawing on the theoretical framework of radicalism by Kruglanski et al. (2014) combined with literature insights on NLP applications to online extremism (Torregrosa et al. 2020, Hitkul et al 2021), it was also possible to establish clear measures of radicalization of a sub-community as well as differences across sub-communities on Parler. The results show that the most popular topics of conversations among Parler communities were centered around Trumpism, conspiracy theories, far-right extremist groups, voter fraud, and conservative in-group favoritism. The texts were subsequently subjected to corpus linguistics tools such as n-gram frequency measures, domain-specific vocabulary significance tests and concordance analysis in order to find patterns in the usage of words. In general, users in radicalized communities deployed a number of strategies in their discourse that is characteristic of perpetuating the far-right narrative thereby exposing more and more people to opportunities for radicalization. This entailed constructing a violent and aggressive discourse with warmongering qualities effectively creating an ‘us vs. them’ dichotomy that legitimizes their need for violence. It was also discovered that radicalized sub-communities used persuasive language in endorsing far-right beliefs to others among their community by leveraging the affordances of social media engagement features as a way to normalize extremist views.
dc.description.sponsorshipUtrecht University
dc.format.extent2347923
dc.format.mimetypeapplication/pdf
dc.language.isoen
dc.titleHow does the language of corpora from radicalized communities discovered on Parler compare to online conversations on Twitter regarding the 2021 Capitol riots and election fraud?
dc.type.contentMaster Thesis
dc.rights.accessrightsOpen Access
dc.subject.keywordsonline extremism, data science, computational linguistics, social media, alt-tech platforms, community detection
dc.subject.courseuuApplied Data Science


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record