FreeSpeech

• 

  home screen

• 

  dashboard-2

• 

  dashboard-1

• 

  login page

Inspiration

Digital classrooms make communication easier, but they also expose students to toxic messages and online bullying. Most schools don’t have real-time moderation tools. I wanted to build something simple, fast, and useful — a system that helps students communicate better and gives teachers clear insights.

What it does

FreeSpeech is an AI-powered chat moderator that:

1. Detects toxic messages instantly
2. Scores severity (low/medium/high)
3. Classifies labels like toxic, insult, identity hate, etc.
4. Stores messages per user
5. Provides a clean analytics dashboard with: a. Toxic vs Safe ratio b. Label distribution c. Toxicity score trends d. Displays Recent harmful messages

It promotes safer, more mindful communication.

How we built it

Backend: Flask + SQLAlchemy Frontend: HTML, CSS, JS, Bootstrap Auth: Flask-Login, hashed passwords ML Model: unitary/toxic-bert Charts: Plotly (Pie, Bar, Line)

Challenges we ran into

1. Making Plotly charts fully responsive (very hectic task)
2. Integrating a transformer model without slowdowns
3. Creating clean UI for both chat and dashboard (conflicts within my own mind)
4. Handling user-specific data safely (safety first)

Accomplishments that we're proud of

1. A complete working product with ML, auth, chat UI, and dashboard
2. Smooth real-time toxicity detection
3. Clean, professional analytics design
4. Real database + actual user behaviour tracking
5. A sense of accomplishment and a bit of confidence boost too.

What we learned

1. Practical Flask architecture
2. Deployable ML integration
3. Visualizing real user data
4. UI/UX for dashboards

What's next for FreeSpeech

1. Toxic word highlighting
2. AI safe rewrite suggestions
3. Teacher/admin panel
4. Weekly behaviour heatmaps

Built With

• bootstrap
• css3
• flask
• html5
• javascript
• machine-learning
• plotly
• pytho
• sqlalchemy