🚀 Release v1.0.0 - Initial Release
Release Date: June 7, 2025
📌 Overview
This initial release launches the Claim Extraction and Normalization System, developed for CLEF-2025 CheckThat! Lab Task 2. The system converts noisy social media posts into clear, normalized claims, significantly enhancing downstream fact-checking workflows.
🌟 Key Features
🎯 Interactive Web Application
-
Real-time claim normalization with streaming AI responses
-
Batch evaluation capability for datasets across various AI models (GPT-4, Claude, Gemini, Llama, Grok)
-
Modern user interface featuring dark mode and real-time evaluation tracking
⚙️ API Backend
-
RESTful endpoints with WebSocket integration for real-time updates
-
Efficient streaming text generation
-
Cross-Origin Resource Sharing (CORS) configured for seamless frontend integration
🧠 Advanced Normalization Techniques
-
Supports multiple prompting methods:
-
Zero-shot & Few-shot prompting
-
Chain-of-Thought (CoT) reasoning
-
Self-Refine iterative refinement
-
Cross-Refine multi-model collaborative refinement
-
Integrated METEOR scoring for automatic claim evaluation and quality assessment
🛠️ Technology Stack
-
Frontend: React, TypeScript, Vite, Tailwind CSS
-
Backend: FastAPI, WebSockets, Streaming Responses
-
AI Model Integration: OpenAI, Anthropic, Google Gemini, xAI Grok, Together.ai
-
Deployment: GitHub Pages (frontend), Render.com (backend)
🚦 Getting Started
-
Quickly set up your local environment using provided automation scripts (setup-project.sh and run-project.sh).
-
Check out the live demo here.
⚠️ Known Issues & Upcoming Roadmap
- Python SDK for API interaction is currently under development.
- Future updates will add Docker containerization, expanded AI model support, and additional evaluation metrics (LLM-as-a-judge).
🤝 Contribution & Licensing
Contributions are highly encouraged!
This project is open-source and licensed under the MIT License.
📄 Full Changelog
Detailed commit history available here.