mikiiiss/Ground_News

AI-Powered Media Analysis is A full-stack AI platform that uses semantic search and NLP to reveal bias, blindspots, and hidden narratives in African news coverage. # 🌍 AfriNews Watch: AI-Powered Media Analysis Platform > **"Breaking the Bubble"**: A full-stack AI platform that uses semantic search and NLP to reveal bias, blindspots, and hidden narratives in African news coverage. --- ## 🚀 The Problem Searching for African news often yields two disconnected realities: 1. **Western Media**: Focuses on conflict, geopolitics, and crisis. 2. **African Media**: Focuses on development, local impact, and national pride. Traditional keyword search fails to connect these dots. It sees "GERD" and "Nile Dam" as different strings, missing the bigger picture. ## 💡 The Solution: Semantic AI **AfriNews Watch** moves beyond keywords. It uses **Vector Embeddings** to understand the *meaning* of articles, grouping them into coherent "Story Clusters" regardless of the specific words used. ### Key Features * **🧠 Semantic Search**: Uses `all-MiniLM-L6-v2` transformer models to find conceptually related stories. * **📊 Bias Detection**: Automatically quantifies the balance of Western vs. African sources for every story. * **🎭 Sentiment Analysis**: NLP-powered "Mood Meters" (Positive/Neutral/Negative) reveal how different regions frame the same event. * **⚠️ Blindspot Alerts**: Warns you when a major story is being ignored by one side of the media divide. * **🔓 AI Internals Mode**: A "Glass Box" feature for engineers that visualizes the raw vector embeddings and model data in real-time. --- ## 🛠️ Technical Architecture ### 1. The "Brain" (ML Engine) * **Embeddings**: Every article is converted into a **384-dimensional vector** using `sentence-transformers`. * **Clustering**: We use **DBSCAN** (Density-Based Spatial Clustering) to dynamically group these vectors into stories without predefined topics. * **Sentiment**: `TextBlob` analyzes the polarity and subjectivity of the content. ### 2. The Backend (FastAPI) * **High-Performance**: Async Python API handling ML inference and data processing. * **Vector Store**: In- …