---
title: Arabic Sentiment Analysis
emoji: 🧠
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.5.1
python_version: '3.10'
app_file: app.py
pinned: false
---


# Arabic Sentiment Analysis with AraBERT


This project presents an **Arabic Sentiment Analysis system** based on a fine-tuned version of **AraBERT**.  
The model predicts the sentiment of Arabic text as **Positive** or **Negative**, along with a confidence score.

The application is deployed using **Gradio** on **Hugging Face Spaces**, providing an interactive web interface for real-time inference.

---

## 🚀 Demo

Simply enter an Arabic text in the input box and the model will return:
- **Sentiment**: Positive / Negative
- **Confidence**: Prediction confidence in percentage (%)

---

## 🧠 Model

- **Base model**: `aubmindlab/bert-base-arabertv02`
- **Task**: Binary Sentiment Classification
- **Labels**:
  - `0` → Negative
  - `1` → Positive
- **Preprocessing**: AraBERT text normalization and cleaning

---

## 📊 Dataset

- **Dataset**: Arabic Sentiment Analysis dataset
- **Classes**:
  - Negative (0): 1,784 samples
  - Positive (1): 2,332 samples
- **Domain**: Arabic tweets and short texts

> ⚠️ The dataset is **not included** in this repository, as it is used only during training.

---

## ⚙️ Training Details

- Fine-tuning performed on **Google Colab**
- **Early Stopping** used to prevent overfitting
- **Checkpointing** enabled during training
- Final model selected based on validation performance

---

## 🖥️ Deployment

- **Framework**: Gradio
- **Platform**: Hugging Face Spaces
- **Hardware**: CPU (Zero GPU)
- **Inference**: Real-time

---

## 📦 Project Structure