ImedBousakhria/darja_bot_sentiment_analysis

This project builds a machine learning model to analyze sentiment in Darja (Algerian Arabic), classifying sentences as positive or negative. It uses a custom dataset containing Darja text in both Arabic script and Latin alphabet. The project applies NLP techniques to handle sentiment classification, tailored specifically for Darja. # Darja Sentiment Analysis This project aims to classify the sentiment of text written in **Darja** (Algerian Arabic) as either **positive** or **negative**. Using natural language processing (NLP) techniques, we have developed a machine learning model tailored to the unique linguistic features of Darja. ## Overview Sentiment analysis for **Darja**, a colloquial form of Arabic spoken in Algeria, presents unique challenges due to its informal structure, diverse vocabulary, and lack of standardized writing. This project tackles these challenges by building a model that identifies whether a Darja sentence expresses **positive** or **negative** sentiment. ## Dataset The dataset used in this project consists exclusively of **Darja** sentences. Each sentence is labeled with a sentiment category: - **Post:** A sentence written in Darja. - **Polarity Class:** The sentiment of the sentence (`0` for negative, `1` for positive). ## Features - **Sentiment Classification:** Classifies Darja text as either positive or negative. - **Preprocessing:** Darja-specific text cleaning, tokenization, and feature extraction using `CountVectorizer`. - **Model Tuning:** Hyperparameter optimization using `MLPClassifier` and `GridSearchCV`. ## Model and Performance - **Model:** `MLPClassifier` - **Performance:** Achieved high accuracy on Darja-only data after model tuning. - **Preprocessing Steps:** - Darja-specific text cleaning - Tokenization and feature extraction with `CountVectorizer` ## Future Work - Expand the dataset to include more Darja sentences for better model generalization. - Explore more advanced NLP models like **transformers** to improve overall accuracy. - use better NER models ## Contributors - Imed Bousakhria- **Imed Bousakhria** - Yassine Cheurfi Belhadj **Cheurfi Behadj Yassine**