---
license: mit
language:
- en
metrics:
- f1
- accuracy
base_model:
- google-t5/t5-base
library_name: transformers
---
# Computational Analysis of Communicative Acts for Understanding Crisis News Comment Discourses

The official trained models for **"Computational Analysis of Communicative Acts for Understanding Crisis News Comment Discourses"**. 

This model is based on **T5-base** and uses the **Compacter** ([Compacter: Efficient Low-Rank Adaptation for Transformer Models](https://arxiv.org/abs/2106.04647)) architecture. It has been fine-tuned on our **crisis narratives dataset**.

---

### Model Information

- **Architecture:** T5-base with Compacter
- **Task:** Single-label classification for communicative act actions
- **Classes:**  
  - `informing statement`  
  - `challenge`  
  - `rejection`  
  - `appreciation`  
  - `request`  
  - `question`  
  - `acceptance`  
  - `apology`  

---

### How to Use the Model

To use this model, you will need the original code from our paper, available here:  
[Acts in Crisis Narratives - GitHub Repository](https://github.com/Aalto-CRAI-CIS/Acts-in-crisis-narratives/tree/main/few_shot_learning/AdapterModel)

#### Steps to Load and Use the Fine-Tuned Model:

1. Add your test task method to `seq2seq/data/task.py`, similar to other task methods.  
2. Modify `adapter_inference.sh` to include your test task's information and this model's name, and then run it.  

```bash
--model_name_or_path CrisisNarratives/adapter-8classes-single_label
```

For detailed instructions, refer to the GitHub repository linked above.

---

### Citation

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

##### TO BE ADDED.

### Questions or Feedback?

For questions or feedback, please reach out via our [contact form](mailto:faezeghorbanpour96@example.com).