Padmavathi Singaravelu@Padmavathi

Overview Our goal is to develop an advanced NLP system that leverages contextual understanding to interpret user queries and provide precise geospatial data. Aadheera is capable of handling both explicit and implicit geospatial information, ensuring that users receive relevant and accurate information in real-time. Key Features: Directions: Provides accurate routes and travel times for seamless navigation. Google Map Integration: Displays interactive maps to enhance visual understanding of routes and locations. Flight Availability: Checks flight availability for effective travel planning. Nearby Services: Finds local services like restaurants and plumbing nearby. Local Events: Discovers local events, keeping users informed about activities in their area. Weather Information: Retrieves current weather details to assist in activity planning. Local News: Fetches relevant local news updates. Location Searches: Searches for specific locations by name, delivering precise data. Technologies Used: NLP Models: Utilizes models like AI71 and Groq for processing queries, fine-tuned for implicit geospatial data extraction. Geospatial Tools & APIs: Google Maps API: Provides geocoding and map visualization. OpenCage Geocode: Offers enhanced geocoding services. SERP API: Supports searches for events, weather, and flights. Database Management: MongoDB: Maintains conversation histories for context-aware responses. Backend Framework: FastAPI: The backend is built using FastAPI, a modern and fast web framework for building APIs with Python. It is deployed on Hugging Face infrastructure, ensuring robust and scalable API management. Frontend Deployment: Vercel: The frontend is designed to display maps and geospatial data visually, enhancing user interaction and experience. It is deployed on Vercel, a platform optimized for frontend frameworks and static sites, ensuring fast and reliable performance.

Our project utilizes Llama 3 and 3.1 to generate synthetic datasets tailored for function calling, with the goal of enhancing small language models in text-to-action tasks. By leveraging these advanced models, we create realistic and varied data that helps train smaller language models to understand and execute commands, queries, and functions more effectively. This approach aims to improve the accuracy and efficiency of language models in interpreting user inputs and performing tasks based on natural language instructions. Ultimately, our project seeks to make language models more capable and reliable in practical applications, enhancing their utility in real-world scenarios.