---
datasets:
- atlasia/No-Arabic-Dialect-Left-Behind
language:
- ar
metrics:
- accuracy
- f1
- recall
pipeline_tag: text-classification
library_name: fasttext
---
# Sfaya-Moroccan-Darija-vs-All

This repository contains a FastText model trained for Moroccan dialect identification. The model achieves a precicion and recall of 0.94 on the atlasia/No-Arabic-Dialect-Left-Behind-v3 test dataset.
It achieves the following performances:

| Label    | F1 Score | Precision Score | Recall Score | False Positive Rate |
|----------|----------|-----------------|--------------|----------------------|
| Morocco  | 0.8986 | 0.9166        | 0.8812     | 0.0129            |
| Other    | 0.9841 | 0.9811        | 0.9871     | 0.1188            |


Try the model using the following:

```python
import fasttext
from huggingface_hub import hf_hub_download

# download model and get the model path
model_path = hf_hub_download(repo_id="atlasia/Sfaya-Moroccan-Darija-vs-All", filename="model_binary_v3_1fpr.bin", cache_dir=None)

# load the model
model = fasttext.load_model(model_path)

# Predict with the model
texts = ["!! رقية بنت محمد"]
predictions = model.predict(texts)

print(predictions)
```

Happy testing! :)