---
license: mit
language:
- ar
metrics:
- accuracy
- f1
- precision
- recall
library_name: transformers
tags:
- offensive language detection
base_model:
- UBC-NLP/MARBERT
---


This model is part of the work done in <!-- add paper name -->. <br>
The full code can be found at https://github.com/wetey/cluster-errors


## Model Details

### Model Description

<!-- Provide a longer summary of what this model is. -->

- **Model type:** BERT-based
- **Language(s) (NLP):** Arabic
- **Finetuned from model:** UBC-NLP/MARBERT

## How to Get Started with the Model

Use the code below to get started with the model.

```python
# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-classification", model="wetey/MARBERT-LHSAB")

```

```python
# Load model directly
from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("wetey/MARBERT-LHSAB")
model = AutoModelForSequenceClassification.from_pretrained("wetey/MARBERT-LHSAB")

```

## Fine-tuning Details

### Fine-tuning Data

This model is fine-tuned on the [L-HSAB](https://github.com/Hala-Mulki/L-HSAB-First-Arabic-Levantine-HateSpeech-Dataset). The exact version we use (after removing duplicates) can be found [](). <!--TODO-->

### Fine-tuning Procedure

The exact fine-tuning procedure followed can be found [here](https://github.com/wetey/cluster-errors/tree/master/finetuning) 

#### Training Hyperparameters

    evaluation_strategy = 'epoch'
    logging_steps = 1,
    num_train_epochs = 5,
    learning_rate = 1e-5,
    eval_accumulation_steps = 2
    
## Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

### Testing Data 

  Test set used can be found [here](https://github.com/wetey/cluster-errors/tree/master/data/datasets)

### Results

`accuracy`: 87.9% <br>
`precision`: 88.1% <br>
`recall`: 87.9% <br>
`f1-score`: 87.9% <br>

 #### Results per class 
| Label | Precision | Recall | F1-score|
|---------|---------|---------|---------|
| normal | 85% | 82% | 83% |
| abusive | 93% | 92% | 93% |
| hate | 68% | 78% | 72% |

## Citation
<!--TODO-->