Categorising keywords: a case study on German conspiracy discourse
Nathan Dykes
Friedrich-Alexander-Universität Erlangen-Nürnberg
Keyword analysis is central to corpus-assisted discourse studies (CADS), as a means of comparing two corpora on a high level. It is typically used to identify starting points for a more detailed analysis.
Usually, keywords are grouped into thematic categories, which are seen as pointers to central topics of the discourse at hand.
However, there is no best practice as to how these categories are formed, and this question has so far received little attention.
In this talk, two different approaches to keyword categorisation in CADS are compared on the keywords of two actors known to spread conspiracies and misinformation on German Telegram channels.
The first strategy examined is the classic approach of topic-based categories, where the categories formed by two independent researchers are compared to explore how individual experts might differ in what central topics are identified.
The second strategy places more focus on linguisic form by annotating surface-level semantic and grammatical features rather than discourse dependent topics.
Overall, the study hopes to open up the discussion with regards to shifting the methodological discussion to the role of the researcher and of linguistic versus thematic categories.