🩺🌐 Health Care Translation Web App

Real-time speech-to-text medical translation, powered by AI

A real-time translation web app built for healthcare communication, combining speech recognition, AI-powered translation, and text-to-speech to bridge language barriers between patients and medical staff.

---

What It Does

• Speech-to-text — captures spoken input directly from the browser microphone
• Real-time translation — translates text via WebSocket communication with low latency
• Text-to-speech — converts the translated text back into spoken audio
• Medical-context aware translation — preserves medical terminology accuracy, cultural appropriateness, and formality (not just literal translation)
• Multi-language support — source and target language selection via dropdowns

---

Branches & Translation Engines

This project explores two different translation backends, implemented on separate branches:

Branch	Translation Engine	Notes
master	OpenAI API (gpt-3.5-turbo)	Context-aware medical translation — understands terminology, tone, and cultural nuance. Requires a paid OpenAI subscription for unrestricted use.
GoogletransAPI	Google Translate API	Faster, free-tier friendly, but lacks the medical-context awareness of the LLM-based approach

The OpenAI-powered branch represents the more advanced implementation — it doesn't just translate words, it adapts medical terminology and tone appropriately for clinical communication.

---

How It Works

User speaks into microphone
        ↓
webkitSpeechRecognition converts speech → text
        ↓
Text + language codes sent via WebSocket (translate_text event)
        ↓
Flask backend receives request → calls OpenAI GPT-3.5-turbo
        ↓
AI translates with medical accuracy + cultural appropriateness
        ↓
translation_response event sent back to client
        ↓
Translated text displayed + converted to speech (SpeechSynthesisUtterance)

---

Tech Stack

Layer	Technology
Backend	Flask (Python)
Real-time communication	Flask-SocketIO
Translation Engine (main)	OpenAI API — gpt-3.5-turbo
Translation Engine (alt branch)	Google Translate API
Speech-to-Text	webkitSpeechRecognition (Web Speech API)
Text-to-Speech	SpeechSynthesisUtterance (Web Speech API)
Frontend	HTML, CSS, JavaScript
Environment Config	python-dotenv
Deployment	Vercel

---

Project Structure

healthcare-translation-app/
│
├── app.py                  # Flask + Socket.IO backend, OpenAI integration
├── requirements.txt        # Python dependencies
├── .env                    # API keys (OPENAI_API_KEY, SECRET_KEY) — gitignored
├── vercel.json              # Vercel deployment configuration
│
├── static/
│   ├── js/
│   │   ├── translation.js  # WebSocket translation requests + UI state handling
│   │   └── speech.js       # Speech-to-text & text-to-speech logic
│   └── css/
│       └── style.css       # UI styling
│
└── templates/
    └── index.html           # Main UI: language selectors, record/translate/speak buttons

---

Core Components

Backend — app.py

• Initializes Flask, Socket.IO, and environment variables
• Calls OpenAI's gpt-3.5-turbo for medically accurate, context-aware translation
• Serves the front-end via / route
• Handles real-time translation requests through the translate_text WebSocket event
• CORS configured to prevent unauthorized cross-origin requests

translation.js

• Sends translation requests over WebSocket
• Listens for translation_response and translation_error events
• Formats translated text and manages UI state (loading spinners, error messages)

speech.js

• Captures spoken input via webkitSpeechRecognition
• Converts translated text to speech via SpeechSynthesisUtterance

---

Security Considerations

• API keys stored securely via .env, excluded from version control via .gitignore
• CORS configured for Socket.IO to prevent unauthorized cross-origin communication
• Backend input validation (e.g. rejecting empty text) to prevent abuse
• Graceful error handling for translation/connection failures
• ⚠️ Rate limiting recommended as a future improvement to protect the OpenAI API from excessive requests

---

Getting Started

Prerequisites

• Python 3.x
• An OpenAI API key (paid subscription recommended — the free trial tier is rate-limited and not suitable for consistent use)

Setup

1. Clone the repository

   git clone https://github.com/sheikhalyan/Health_Care.git
   cd HealthCare

2. Create a virtual environment and install dependencies

   python -m venv venv
   source venv/bin/activate  # Windows: venv\Scripts\activate
   pip install -r requirements.txt

3. Create a .env file in the root directory

   OPENAI_API_KEY=your_openai_api_key_here
   SECRET_KEY=your_flask_secret_key_here
   

4. Run the app

   python app.py

5. Open http://localhost:5000 in your browser

---

Switching Translation Engines

To try the Google Translate API implementation instead of OpenAI:

git checkout GoogletransAPI

This branch swaps the OpenAI translation call for the Google Translate API — useful for comparing cost, latency, and translation quality between an LLM-based approach and a traditional translation API.

---

⚠️ Disclaimer

This is a prototype built for demonstration and learning purposes. It is not a certified medical communication tool and should not be relied upon for critical clinical interactions without further validation and professional medical review.

---

Author

Sheikh Alyan — Full-Stack Developer