ABSTRACT
This section introduces relational methods that are for extracting topics from a text corpus. These methods stem from two different paradigms: geometric data analysis and semantic network analysis. Both offer unique advantages. On the one hand, geometric data analysis enables us to integrate multiple data types in a single analysis and allows us to map topics in relation to document metadata. Geometric data analysis encompasses Correspondence Analysis, Multiple Correspondence Analysis, Principal Component Analysis, and Multiple Factor Analysis, allowing for efficient analysis and visualization of different data types. On the other hand, semantic network analysis captures local meaning structures and quantifies them using centrality measures and community detection algorithms. A key advantage of semantic network analysis is that these local meaning structures are embedded within the broader network structure, which allows us to simultaneously track the evolution of topics and the discourse structure over time.
