# Argument–Keypoint Matching with DistilBERT

This model predicts whether an argument is correctly matched (**Apparié**) or not (**Non-Apparié**) with a given key point.


## Model Description
- **Base Model:** DistilBERT (uncased)
- **Task:** Binary text-pair classification
- **Training Data:** [IBM ArgKP-2023 dataset (~32,000 examples)](https://research.ibm.com/haifa/dept/vst/debating_data.shtml)
- **Labels:**
  - `0` — Non-Apparié
  - `1` — Apparié
- **Input:** (argument, key_point)
- **Output:** Predicted class + probabilities


## Performance
- Strong accuracy and F1 score on evaluation data
- Reliable predictions across both labels



## Training

Trained on a balanced argument–keypoint dataset

Exported using save_pretrained



## Citation

```
@misc{argument-keypoint-matching,
  author       = {Malek Messaoudi},
  title        = {Argument–Keypoint Matching with DistilBERT},
  year         = {2025},
  publisher    = {Hugging Face},
  howpublished = {{\\url{{https://huggingface.co/NLP-Debater-Project/destlibert-keypoint-matching}}}}
}

````

## License

MIT License