---
license: mit
base_model:
- deepset/gbert-base
pipeline_tag: text-classification
language:
- de
---


This model is a fine-tuned version of [deepset/gbert-base](https://huggingface.co/deepset/gbert-base). It was fine-tuned on the task of text classification for conspiracy theory detection.
The model was retrained on ~ 1.25 million German-language messages from conspiracist Telegram channels, and then fine-tuned using the dataset TelCovACT dataset, which comprises ~ 4.000 German-language Telegram message annotated for conspiracy theories. The TelCovACT dataset is available upon request. For more information see the [dataset paper](https://ojs.aaai.org/index.php/ICWSM/article/view/22216) and the [datasheet](https://zenodo.org/records/7093870).
For fine-tuning details and evaluation see the [reference paper](https://arxiv.org/abs/2404.17985).

## Citation Information
Please cite the reference paper if you use this model:

```bibtex
@misc{pustet2024detection,
  title={Detection of Conspiracy Theories Beyond Keyword Bias in German-Language Telegram Using Large Language Models},
  author={Milena Pustet and Elisabeth Steffen and Helena Mihaljević},
  year={2024},
  eprint={2404.17985},
  archivePrefix={arXiv}
}
```