BhashaSuraksha

Preserving India's Linguistic Heritage through AI

BhashaSuraksha is an open-source platform designed to document, preserve, and analyze endangered and low-resource Indian languages. It leverages advanced speech processing, machine learning, and geospatial visualization to create a living digital archive of India's diverse linguistic landscape.

Features

• Audio Collection: Secure upload and recording interface for gathering speech samples from diverse regions.
• AI-Powered Analysis:
  • Transcription: Automatic speech-to-text using OpenAI Whisper.
  • Dialect Identification: Acoustic embedding extraction using wav2vec2.
  • Semantic Analysis: Keyword extraction and cultural context analysis using Google Gemini AI.
• Unsupervised Clustering: Automatically groups unknown or undocumented dialects using DBSCAN and Cosine Similarity on acoustic embeddings.
• Geospatial Heatmap: Interactive visualization of language distribution and dialect clusters using MapLibre.
• Scalable Architecture: Built on Azure Blob Storage and PostgreSQL for robust data handling.

Tech Stack

Frontend

• Framework: Next.js 15 (React)
• Styling: Tailwind CSS, Framer Motion
• Maps: MapLibre GL JS
• State Management: React Hooks

Backend (ML Service)

• Framework: FastAPI (Python)
• ML Models:
  • openai/whisper (Transcription)
  • facebook/wav2vec2-large-xlsr-53 (Embeddings)
  • scikit-learn (Clustering)
• Processing: Librosa, NumPy, PyTorch

Infrastructure & Data

• Database: PostgreSQL (with Prisma ORM)
• Storage: Azure Blob Storage
• AI Services: Google Gemini API
• Containerization: Docker

Architecture

The system consists of two main microservices:

1. Frontend (Next.js): Handles user interaction, audio recording, and visualization. It communicates directly with Azure for uploads and the ML service for processing.
2. ML Service (FastAPI): Processes audio files to extract transcripts and embeddings, performs clustering, and manages the vector database.

Getting Started

Prerequisites

• Node.js 18+
• Python 3.10+
• PostgreSQL
• Azure Storage Account
• Google Gemini API Key

Installation

1. Clone the repository

   git clone https://github.com/Twahaaa/bashasuraksha.git
   cd bashasuraksha

2. Setup Frontend

   cd frontend
   npm install
   
   # Create .env file
   cp .env.example .env.local
   # Update .env.local with your credentials
   
   npm run dev

3. Setup ML Service

   cd ml
   python -m venv venv
   source venv/bin/activate  # or venv\Scripts\activate on Windows
   pip install -r requirements.txt
   
   # Create .env file
   cp .env.example .env
   # Update .env with your credentials
   
   uvicorn app.main:app --reload --port 8000

Environment Variables

Frontend (.env)

DATABASE_URL="postgresql://..."
NEXT_PUBLIC_AZURE_STORAGE_SAS_URL="..."
GEMINI_API_KEY="..."

ML Service (.env)

DATABASE_URL="postgresql://..."
AZURE_STORAGE_CONNECTION_STRING="..."

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License.