Bhashini Live Website Translation Platform
Real-time website translation platform for Bhashini (MeitY) with a Maker-Checker quality assurance workflow for human translators.
Project Snapshot
Role
Lead Developer
Organization
Bhashini (Ministry of Electronics & IT, Government of India)
Timeline
2025
Status
Live
Deployment
Production - deployed on Bhashini AI Cloud Infrastructure
Tech Stack
Impact & Results
- Live production deployment - used by government departments
- Supports all major Indian languages via Bhashini models
- Human-in-the-loop quality verification system
A real-time website translation platform built for Bhashini (MeitY) that includes a Maker-Checker quality assurance workflow to ensure government-standard translations using human-in-the-loop verification.
The Problem
Bhashini needed a platform to: - Enable live/real-time translation of any website into any Indian language. - Provide a Maker-Checker quality assurance system for human translators to verify and correct AI translations, ensuring government-standard translation quality.
The Solution
A full-stack translation platform where browser-side scripts capture webpage text, route it to a Django backend for AI translation, and serve it back to the page. It includes a dashboard for human translators to review and refine the output.
Architecture & Implementation
Data Capture: Browser-side script/plugin captures webpage text content.
Translation Pipeline: Sends content to the Django backend for translation via Bhashini's translation APIs, rendering it back on the page in the target Indian language.
Quality Assurance: A Maker-Checker workflow where human translators review AI-translated segments, approve or correct them, and the system learns from these corrections to maintain standards.
Monitoring: A dedicated translation quality monitoring dashboard.
Data Used: Live website text and translations across all major Indian languages.
Tech Stack: Python, Django, React.js, Bhashini AI Cloud, Translation APIs, PostgreSQL.
Challenges & Solutions
DOM Integrity: Preserving the original DOM structure and styles while replacing text strings dynamically.
UI/UX: Designing a seamless maker-checker UI for human translators that integrates with the AI output.
Scale: Handling translation requests for entire websites efficiently.
Key Takeaways
Complex DOM manipulation and browser-side scripting.
Efficient Django backend routing and API design.
Building robust human-in-the-loop AI verification workflows.