Skip to content

SpeakSafe is a real-time voice-to-text transcription app with profanity detection and filtering. It ensures clean, safe, and accessible speech content for all users (WIP)

License

Notifications You must be signed in to change notification settings

swet40/SpeakSafe

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SpeakSafe

Real-time fraud call detection system built with Flask, React, and Socket.IO. It leverages machine learning (using scikit-learn and joblib) to analyze call patterns and detect fraudulent activity instantly. The frontend, developed in React, provides a seamless user experience, while Flask and Socket.IO ensure real-time communication and detection.

Features

-Real-time fraud detection using audio transcription and AI models. -Socket.IO integration for continuous streaming and real-time processing. -Machine Learning model trained on fraudulent and non-fraudulent messages. -Deepgram API for accurate speech-to-text transcription. -REST API and WebSocket support for seamless integration.

Technologies Used

Frontend: React.js (Run using npm install && npm start) -Backend: Flask, Flask-SocketIO (Run using pip install -r requirements.txt && python app.py) -ML & AI: Scikit-Learn, Google Gemini AI, Deepgram API -Other: dotenv, joblib, requests, numpy, pandas, nltk

Installation

Backend Setup

-Clone the repository :git clone https://github.com/swet40/SpeakSafe.git -Install Dependicies : pip install -r requirements.txt

  • Set up environment variables: GOOGLE_API_KEY=your_gemini_api_key DEEPGRAM_API_KEY=your_deepgram_api_key CORS_ALLOWED_ORIGIN=*
  • Run the Flask Server: python app.py

Frontend Setup

  • Navigate to the frontend folder : cd Frontend
  • Install Dependicies: npm install
  • npm start

How It Works

  • Audio Streaming: The frontend sends audio data in chunks to the backend via Socket.IO.
  • Real-Time Transcription: The Deepgram API transcribes each chunk as it arrives.
  • NLP Processing: The transcribed text is vectorized and analyzed using an ML model.
  • Fraud Detection: Gemini AI analyzes the text and determines fraud probability with reasoning.
  • Immediate Response: The backend sends fraud detection results for each audio chunk back to the frontend in real time.

Dataset

  • The ML model is trained using a dataset containing labeled fraud and normal text messages.

About

SpeakSafe is a real-time voice-to-text transcription app with profanity detection and filtering. It ensures clean, safe, and accessible speech content for all users (WIP)

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published