SpeechFlowguard

🔗 GitHub Link: SpeechFlowguard

Tech Stack

• Python
• FastAPI
• Scikit-Learn
• TF-IDF Vectorizer
• Logistic Regression
• Natural Language Processing (NLP)
• Docker

Features

Real-Time Toxicity Detection

• Analyzes user comments in real-time to detect toxic content.
• Handles multi-label classification for categories:
  • toxic
  • severe_toxic
  • obscene
  • threat
  • insult
  • identity_hate

Machine Learning Pipeline

• Uses TF-IDF for feature extraction.
• Trained Logistic Regression models for each label.
• Lightweight and interpretable models for fast inference.

Web API Interface

• Built using FastAPI for high-performance async handling.
• REST endpoints allow sending a comment and receiving predictions instantly.

Dockerized Deployment

• Fully containerized using Docker.
• Easily deployable as a microservice on cloud or edge environments.

Workflow

1. Model Training
   • Train multi-label classifiers using scikit-learn and TF-IDF on annotated datasets.
2. API Development
   • Create endpoints in FastAPI for:
     • /predict: Accepts user comment and returns predicted labels.
3. Containerization
   • Use Docker to package the model and API into a modular image.
4. Deployment
   • Deployable on any cloud or local server with Docker support.

Highlights

• 🚀 Real-time inference with low latency.
• 🧠 Simple, fast, and interpretable ML architecture.
• 🧱 Modular codebase for easy retraining or replacement of components.
• ☁️ Cloud-ready via Docker.