---
license: mit
language:
- de
base_model:
- deepset/gbert-large
---
# Autor-Regulatory Focus Classifier (German)

This model is a fine-tuned transformer-based classifier that detects the **regulatory focus** in German-language text, classifying whether the language expresses a **promotion** (aspirational, growth-oriented) or **prevention** (safety, obligation-oriented) focus.

It is fine-tuned on top of a German-language base model for the task of binary text classification.

## Model Details

- **Base model**: `deepset/gbert-large` 
- **Fine-tuned for**: Binary classification (Regulatory Focus)
- **Language**: German
- **Framework**: Hugging Face Transformers
- **Model format**: `safetensors`

## Use Cases

- Social psychology and communication research
- Marketing and consumer behavior analysis
- Literary or political discourse analysis
- Behavioral modeling and goal orientation profiling

## Example Usage

```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model = AutoModelForSequenceClassification.from_pretrained("aveluth/author_regulatory_focus_classifier")
tokenizer = AutoTokenizer.from_pretrained("aveluth/author_regulatory_focus_classifier")

text = "Wir müssen sicherstellen, dass keine Fehler passieren. Sicherheit hat höchste Priorität."
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)
predicted_class = torch.argmax(outputs.logits).item()

print("Predicted class:", "prevention" if predicted_class == 0 else "promotion")
```

## Labels

| Class       | Description                            |
|-------------|----------------------------------------|
| `0`         | Prevention-focused language            |
| `1`         | Promotion-focused language             |

## Training Details

- **Training data**: Custom labeled corpus based on psychological framing
- **Loss function**: Cross-entropy
- **Optimizer**: AdamW
- **Epochs**: 4
- **Learning rate**: 3e-5

## Limitations

- Trained on German-language data only
- Performance may vary on out-of-domain text (e.g., technical manuals, poetry)
- May not generalize across all cultural framings of regulatory focus

## License

[MIT](LICENSE)

## Citation

If you use this model in your research, please cite:

```bibtex
@article{velutharambath2023prevention,
  title={Prevention or Promotion? Predicting Author's Regulatory Focus},
  author={Velutharambath, Aswathy and Sassenberg, Kai and Klinger, Roman},
  journal={Northern European Journal of Language Technology},
  volume={9},
  number={1},
  year={2023}
}
```