HichamDe/darija-offensive-language-detection

# 🚀 Offensive Message Detection in Arabic Darija This project focuses on building a **classification model** to determine whether a message is **offensive** or **not** in **Arabic Darija**. The model has practical applications, such as: ✅ **Filtering offensive comments** ✅ **Detecting hate speech & harmful content** --- ## 📌 Project Steps ### 🔍 1. Data Gathering - 📥 Collected a **high-quality dataset** with labeled offensive and non-offensive messages. - 🔎 Found an existing **Data set** online for reference. ### 📊 2. Exploratory Data Analysis (EDA) - Used **Pandas** to analyze and clean the dataset. - Identified missing values, text patterns, and class imbalances. ### 🏗 3. Feature Engineering - Converted text data into numerical form using **Vectorization (TF-IDF / CountVectorizer)**. ### 🎯 4. Model Selection - Selected **Logistic Regression** as the classification model. ### ✂️ 5. Splitting Data - Split the dataset into **training (80%)** and **testing (20%)** for evaluation. ### 📈 6. Model Evaluation - Measured accuracy, precision, recall, and F1-score to validate performance. ### 🛠 7. Running Custom Tests - Tested the model with **real-world Arabic Darija messages** to assess its accuracy. --- ## 🚀 Future Improvements 🔹 Train with a **larger dataset** for better accuracy. 🔹 Experiment with **deep learning models (LSTMs, Transformers)** for improved results. 🔹 Deploy the model as an **API** for real-time detection. --- ## 🛠 Tech Stack Used - 🐍 **Python** - 📊 **Pandas, NumPy** - 🤖 **Scikit-learn** - 📝 **NLTK / Text Preprocessing Libraries** --- ## 📢 Contributions & Feedback Feel free to contribute, suggest improvements, or test the model! 🚀