Inspiration
Competitive debating demands quick thinking, structured reasoning, and constant improvement. However, many debaters lack access to high-quality feedback outside of formal coaching or tournaments. We wanted to create a tool that democratizes debate coaching—something that can analyze, evaluate, and help improve argumentation in real time.
What it does
DialectIQ is a real-time cognitive coach for debaters. Using speech-to-text and AI-powered argument analysis, it listens to a live debate, identifies key claims, logical structures, fallacies, emotional tones, and rhetorical strategies. It then provides immediate, constructive feedback—highlighting strengths, weaknesses, and suggested improvements. It also tracks metrics over time such as clarity, pace, logic, and persuasiveness.
How we built it
We used:
Whisper for accurate real-time speech-to-text transcription
OpenAI’s GPT-4o to perform natural language understanding, logical reasoning analysis, and coaching feedback
A React frontend for the live debate interface and dashboard
Flask backend to handle API calls and data management
Socket.io for real-time communication between the frontend and backend
MongoDB for storing user data, feedback history, and performance metrics
Challenges we ran into Ensuring real-time performance without compromising accuracy
Managing latency between transcription, analysis, and feedback delivery
Designing feedback that is helpful but non-intrusive during live debates
Training GPT to recognize formal argumentation structures consistently
Handling diverse speaking styles, accents, and emotional tones
Accomplishments that we're proud of
Achieved sub-3 second feedback latency from live speech
Successfully identified logical fallacies like ad hominem, false dichotomy, and red herring in real-time
Built a clean, intuitive UI for both debaters and coaches
Developed a feedback engine that balances critique with encouragement
What we learned
Real-time NLP is extremely context-sensitive and requires careful optimization
Debate evaluation involves more than correctness—it’s also about persuasion, confidence, and structure
Users respond best to coaching that combines data-driven feedback with human-like tone and empathy
What's next for DialectIQ – Real-Time Cognitive Coach for Debaters Expanding support for multiple debate formats (e.g., British Parliamentary, Lincoln-Douglas, Policy)
Introducing team-based feedback aggregation and strategy suggestions
Creating personalized improvement plans based on user history
Adding video analysis to assess body language and visual rhetoric
Partnering with schools and debate organizations for pilot programs
Log in or sign up for Devpost to join the conversation.