Masterarbeit, Otto-Friedrich-Universität Bamberg, 2026

This thesis investigates how the semantics of political terms vary across ideological contexts using methods from distributional semantics. The analysis is based on word embeddings trained with Word2Vec on a corpus of parliamentary speeches from the German Bundestag and the European Parliament, covering the period from 2017 to 2025. The corpus is divided into left-wing, centrist, and right-wing subsets.

To interpret these semantic differences, the study draws on Schwartz Value Theory as a theoretically grounded framework. By relating the embeddings of political terms to the embeddings of Schwartz values, the analysis examines whether semantic differences align with underlying ideological value structures.

The results show that semantic variation is not uniform. Institutional and policy-related terms remain relatively stable, while culturally loaded terms exhibit stronger ideological divergence. At the same time, the expected alignment between value associations and ideological positions is not supported, suggesting that the relationship between language use and underlying value structures is more complex and context-dependent than initially assumed.