---
library_name: transformers
license: mit
base_model: microsoft/mdeberta-v3-base
tags:
- text-classification
- hierarchical-classification
- climate-change
- ukraine-russia-war
- narrative-classification
- multilingual
datasets:
- custom
language:
- en
- bg
- hi
- pt
- ru
metrics:
- f1
- precision
- recall
pipeline_tag: text-classification
---

# mDeBERTa v3 Base - Hierarchical Narratives Classifier

This model is a fine-tuned version of [microsoft/mdeberta-v3-base](https://huggingface.co/microsoft/mdeberta-v3-base) for hierarchical text classification of narratives related to Climate Change and Ukraine-Russia War.

## Model Description

This model performs hierarchical classification on text documents to identify:
1. **Main Narrative Categories**: Climate Change (CC) and Ukraine-Russia War (URW) related narratives
2. **Sub-Narrative Classifications**: Detailed subcategories within each main narrative

## Supported Languages

The model has been trained and tested on multiple languages:
- English (EN)
- Bulgarian (BG) 
- Hindi (HI)
- Portuguese (PT)
- Russian (RU)

## Usage

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

# Load model and tokenizer
model_name = "AWCO/mdeberta-v3-base-narratives-classifier-hierarchical"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModel.from_pretrained(model_name)

# Example usage for inference
text = "Your text here..."
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)

with torch.no_grad():
    outputs = model(**inputs)
    # Process outputs according to your inference pipeline
```

## Training Data

The model was trained on a custom dataset containing narratives related to:
- Climate Change discussions and misinformation
- Ukraine-Russia War narratives and propaganda

## Performance

The model achieves strong performance across multiple languages for both narrative-level and sub-narrative level classification tasks.

## Model Files

- `config.json`: Model configuration
- `model.safetensors`: Model weights in SafeTensors format
- `tokenizer.json`: Tokenizer configuration
- `label_mappings_hierarchical.json`: Hierarchical label mappings for classification

## Citation

If you use this model in your research, please cite accordingly.

## License

This model is released under the MIT License.