The analysis builds on recent advances in Large Language Models (LLMs) and computational text analysis to process and interpret large-scale parliamentary and social media corpora. These corpora are designed to be comparable across countries, which makes it possible to explore far-right topics cross-linguistically, examining how they are referred to and framed according to political orientation. Two complementary approaches are integrated: (a) a keyword-based retrieval pipeline, designed to capture targeted instances of relevant discourse phenomena, and (b) topic modeling, employed to automatically identify recurring themes, semantic clusters, and linguistic patterns across texts. LLMs are used both for the semantic expansion of keyword queries and for generating contextual embeddings that enhance the interpretability and coherence of topics, as well as for evaluating the coherence and validity of the topics identified. This combined strategy bridges hypothesis-driven and exploratory analysis, enabling systematic cross-linguistic and cross-platform comparisons, as well as the examination of temporal dynamics in online communication. The resulting framework offers a reproducible, multilingual, and scalable methodology for analysing large-scale textual data from parliamentary corpora and social media sources.
