Researchers develop guidelines and corpus for extracting conspiratorial actors from German Telegram posts

The authors address the lack of computational research on identifying specific actors in conspiracy theories by developing annotation guidelines and a span-annotated corpus of German Telegram posts. They investigate the automatic extraction of these actors using transformer-based models and apply the resulting model to the Schwurbelarchiv archive.

• The study presents new annotation guidelines specifically for conspiratorial actors.
• A span-annotated corpus of German Telegram posts is introduced for this purpose.
• Transformer-based models are used to automatically extract these actors from the text.
• The approach is applied to the Schwurbelarchiv, a large-scale archive of German conspiracy-related Telegram channels.

The results demonstrate that conspiratorial actors can be annotated with meaningful agreement and extracted with reasonable accuracy despite linguistic complexity, enabling large-scale analyses of actor representations in conspiracy narratives.