|
---
|
|
language: ar
|
|
tags:
|
|
- arabic
|
|
- sentiment-analysis
|
|
- Farasa
|
|
- AraBERT
|
|
datasets:
|
|
- ArSAS
|
|
license: apache-2.0
|
|
---
|
|
|
|
# Arabert Sentiment Model with Farasa Preprocessing
|
|
|
|
This is a fine-tuned version of `AraBERT` using the ArSAS dataset for **sentiment analysis**.
|
|
The text was preprocessed using **Farasa** for optimal tokenization of Arabic text.
|
|
|
|
## Model Details
|
|
|
|
- **Base Model**: AraBERT v2
|
|
- **Dataset**: ArSAS (Arabic Sentiment Analysis)
|
|
- **Preprocessing**: Farasa Tokenization
|
|
- **Tasks**: Sentiment Classification (`negative`, `neutral`, `positive` , `mixed`)
|
|
|
|
## Usage
|
|
|
|
```python
|
|
from transformers import pipeline
|
|
|
|
sentiment_pipeline = pipeline(
|
|
task="text-classification",
|
|
model="Abdo36/Arabert-Sentiment-Analysis-ArSAS"
|
|
)
|
|
|
|
result = sentiment_pipeline("هذا المنتج رائع للغاية")
|
|
print(result)
|
|
```
|
|
|
|
## Training Details
|
|
|
|
- **Number of Epochs**: 2
|
|
|
|
- **Training Loss**: 0.6
|
|
|
|
- **Validation Accuracy**: 0.5
|
|
|
