Musajidda/Hausa-Sentiment-Analysis

# 📘 Improved Hausa Sentiment Analysis This project focuses on enhancing Aspect-Based Sentiment Analysis (ABSA) for the Hausa language using a Deep Learning approach combining Convolutional Neural Networks (CNN) and Attention Mechanism. It addresses the challenges of low-resource language processing and supports multi-label aspect classification and sentiment polarity classification specifically in Hausa-language movie reviews. # 📌 Project Highlights Language Focus: Hausa (spoken by over 50 million people in West Africa) Model Architecture: CNN + Attention Tasks: Multi-label Aspect Classification (e.g., acting, storyline, character) Single-label Sentiment Classification (positive, negative, neutral) Dataset: Hausa movie reviews with manually annotated aspects and sentiments Use Case: Useful for media monitoring, digital marketing, social media opinion mining, and regional NLP research. # 🔧 Features 🧠 Deep Learning with Keras & TensorFlow 🗂️ TF-IDF Vectorization for text preprocessing 🎯 Evaluation Metrics: Accuracy, Precision, Recall, F1-score 🌍 Support for under-resourced African languages 📊 Visual analysis and performance reports 🚀 Getting Started Clone the repo git clone github.com cd hausa-sentiment-analysis # 📄 Research Paper This project is based on an academic research titled: "Improved Aspect-Based Sentiment Analysis for Hausa Movie Reviews using CNN + Attention" It extends a baseline model (Ibrahim et al., 2024) to support multi-aspect classification. # 💡 Future Plans Expand the dataset to include more aspect categories Experiment with Transformer-based models like mBERT or AfriBERTa Deploy a web-based sentiment analysis demo # 🤝 Acknowledgements Special thanks to the HausaNLP community and researchers who contributed to low-resource NLP development. ## ✍️ Author **Mohammed Musa Jidda** Researcher | NLP Enthusiast | Developer 📧 Email: muhammadmjidder8@gmail.com GitHub Profile