🛡️ MYTH CHASER

Anti-scam and myth-busting utility powered by AI

MYTH CHASER is an advanced AI-powered fact-checking and anti-scam detection system that helps users verify the authenticity of claims, statements, and media content. Using multiple machine learning models and real-time web search capabilities, it provides comprehensive analysis to identify facts, myths, and potential scams.

✨ Features

🤖 Multi-Model AI Analysis

• Natural Language Inference (NLI): RoBERTa-based model for logical reasoning between claims and evidence
• Sentence-BERT (SBERT): Semantic similarity analysis using sentence embeddings
• ClaimBuster Integration: Professional fact-checking API for claim verification
• Google Fact Check API: Access to Google's comprehensive fact-checking database
• Fake News Detection: Specialized model for identifying misinformation patterns
• Groq Qwen3-32B: Advanced large language model with sophisticated reasoning capabilities
• TunBERT: Specialized Arabic and Tunisian dialect fact-checking model
• AI Explanation Generator: Groq-powered detailed explanations for all verdicts

🔍 Multi-Format Content Analysis

• Text Processing: Direct text input analysis and verification
• Image Analysis: OCR text extraction and visual content description using BLIP
• Audio Processing: Speech-to-text conversion for audio content verification
• Drag & Drop Interface: Seamless file upload with support for multiple formats

🌐 Multi-Language Support

• Auto-Detection: Automatic language identification for incoming content
• Translation Engine: Google Translate integration for seamless cross-language analysis
• Arabic Dialect Support: Specialized handling for Tunisian Arabic and transliterated text
• Language Preservation: Original text maintained for dialect-specific models

🌐 Real-Time Web Search

• DuckDuckGo Integration: Automated web search for evidence gathering
• Source Aggregation: Intelligent collection and processing of relevant information
• Evidence Synthesis: Combines multiple sources for comprehensive analysis

🎯 Three-Tier Classification System

• FACT: Verified true statements with supporting evidence
• MYTH: Partially true or misleading information requiring clarification
• SCAM: False, harmful, or deceptive content

💡 Advanced Weighted Voting Algorithm

• Multi-Model Consensus: Combines predictions from 7 specialized AI models
• Intelligent Weighting: Groq Qwen3-32B receives highest voting power (3x weight)
• Confidence-Based Filtering: Ignores uncertain predictions for cleaner results
• Graceful Degradation: System continues functioning even with individual model failures

🏗️ Architecture

Backend (FastAPI)

• Multi-threaded Processing: Parallel execution of AI models for faster analysis
• RESTful API: Clean /classify endpoint for content verification
• CORS Support: Seamless frontend-backend communication
• Error Handling: Robust error management and graceful degradation

Frontend (Next.js 15)

• Modern React Architecture: Built with React 19 and Next.js 15
• TypeScript Support: Full type safety throughout the application
• Tailwind CSS: Responsive, modern UI design
• Real-time Feedback: Loading states and progress indicators
• File Management: Advanced file upload with preview and management

Machine Learning Models

• Transformer-based: State-of-the-art NLP models for text analysis
• Large Language Models: Groq Qwen3-32B for advanced reasoning and explanation
• Specialized Dialects: TunBERT for Arabic and Tunisian language support
• Computer Vision: BLIP model for image understanding
• Speech Recognition: Google Speech Recognition for audio processing
• Ensemble Methods: Multiple model predictions combined for accuracy

🛠️ Technology Stack

Backend

• FastAPI: High-performance Python web framework
• PyTorch: Deep learning framework for AI models
• Transformers: Hugging Face transformers library
• Groq: Advanced LLM API integration
• SpeechRecognition: Audio processing capabilities
• PIL/Pytesseract: Image processing and OCR
• GoogleTrans: Multi-language translation support
• DuckDuckGo Search: Web search integration

Frontend

• Next.js 15: React-based web framework
• React 19: Latest React with concurrent features
• TypeScript: Type-safe development
• Tailwind CSS: Utility-first CSS framework
• Custom Hooks: Reusable state management logic

AI Models

• ynie/roberta-large-snli_mnli_fever_anli_R1_R2_R3-nli: NLI classification
• all-MiniLM-L6-v2: Sentence embeddings
• winterForestStump/Roberta-fake-news-detector: Fake news detection
• not-lain/TunBERT: Arabic and Tunisian dialect fact-checking
• Groq Qwen3-32B: Advanced reasoning and explanation generation
• Salesforce/blip-image-captioning-base: Image captioning

📁 Project Structure

Hack-AINS/
├── apis/                     # Backend API server
│   ├── main.py              # FastAPI application entry point
│   ├── models/              # AI model implementations
│   │   ├── NLI/             # Natural Language Inference
│   │   ├── SBERT/           # Sentence-BERT similarity
│   │   ├── ClaimBuster/     # ClaimBuster API integration
│   │   ├── Google/          # Google Fact Check API
│   │   ├── FakeNewsDetector/# Fake news classification
│   │   └── Explainer/       # AI explanation generation
│   ├── converters/          # Media-to-text conversion utilities
│   ├── web_searcher/        # Web search and evidence gathering
│   └── requirements.txt     # Python dependencies
│
├── frontend/                # Next.js web application
│   ├── app/                 # Next.js app directory
│   │   ├── components/      # React components
│   │   ├── hooks/           # Custom React hooks
│   │   ├── types/           # TypeScript type definitions
│   │   └── utils/           # Utility functions
│   └── package.json         # Node.js dependencies
│
└── compose.yaml             # Docker composition (future deployment)

🚀 Key Capabilities

• Real-time Analysis: Process and verify content within seconds
• Multi-modal Support: Handle text, images, and audio files seamlessly
• Source Verification: Cross-reference claims with multiple authoritative sources
• Confidence Scoring: Provide reliability indicators for all predictions
• Batch Processing: Handle multiple files simultaneously
• Responsive Design: Works across desktop and mobile devices
• Privacy-focused: No data retention, immediate processing and disposal

🎨 User Interface

• Retro Gaming Aesthetic: Pixel-perfect design with nostalgic appeal
• Drag & Drop: Intuitive file upload with visual feedback
• Real-time Feedback: Loading animations and progress indicators
• Color-coded Results: Green (FACT), Orange (MYTH), Red (SCAM)
• Detailed Explanations: AI-generated reasoning for all verdicts
• File Management: Preview, manage, and remove uploaded files

---

MYTH CHASER - Fighting misinformation with artificial intelligence 🤖✨