Framing Extremist Narratives: A Multi‐Stage Computational Analysis of Discourse Structures, Emotional Mobilization, and Cross‐Platform Migration

ABSTRACT

Online extremism poses a particularly difficult governance challenge when radical actors adapt and spread grievance, authority, enemy, and recruitment narratives across heterogeneous platforms. This study develops a multi‐stage computational framework for analyzing extremist narratives across platforms. Using a corpus of 22,347 English‐language documents from Twitter/X, Telegram, Gab, public web archives, and extremist‐adjacent forums posted between 2020 and 2023, the study integrates topic modeling, frame classification, psycholinguistic and emotion scoring, semantic cross‐platform matching, qualitative validation, and negative‐control checks. The findings identify eight topic clusters, six recurring narrative frames, and a platform‐migration pattern associated with semantic drift and stronger enemy‐focused framing in less‐moderated destination spaces. The results are consistent with a displacement‐and‐intensification mechanism, although they do not establish causal effects of deplatforming or individual radicalization. The study contributes to computational discourse analysis, coordinated platform governance, and frame‐aware counter‐narrative design.