CEREBRO — Neural Sentiment Hackathon
Where minds decode emotion at machine speed.
CEREBRO is a 36-hour online hackathon organized by the Department of Computer Science & Engineering, Institute of Engineering & Management (IEM), Kolkata.
The hackathon challenges participants to build systems that can understand and analyze human emotion, sentiment, and tone across different forms of communication.
Participants will choose one of three problem statements and develop a functional solution using their own technology stack, models, and approach.
CEREBRO focuses on practical applications of Artificial Intelligence, Machine Learning, Natural Language Processing, Speech Processing, and Affective Computing.
Problem Statements
PS-01 — Tone Intelligence in Conversations
First Prize: ₹5,000
Chat conversations contain more than words. They can reveal tension, sarcasm, warmth, frustration, and escalation.
Build a system that analyzes the emotional arc of a multi-turn conversation and identifies changes in tone throughout the conversation.
Key requirements:
Multi-turn chat sentiment and tone tracking
Sarcasm, passive-aggression, and irony detection
Emotion-arc visualization over the conversation
Support for chat formats such as WhatsApp, Slack, Discord, or similar exports
Participants must source their own data and develop their own approach
PS-02 — Voice as a Mood Signal
Second Prize: ₹3,000
The emotional state behind a sentence can change dramatically depending on how it is spoken.
Build a system that extracts emotional information directly from raw audio, using characteristics such as pitch, prosody, speech rate, and pauses.
Key requirements:
Audio-based emotion classification
Continuous valence and arousal estimation
Robustness to noisy, accented, or code-switched speech
Emotion should primarily be inferred from the audio signal itself
Participants must source their own training data
PS-03 — Sentiment Drift in Social Threads
Third Prize: ₹2,000
Online discussions can begin calmly and quickly become hostile or toxic.
Build a system that analyzes a sequence of comments and identifies where the emotional tone changes, including the specific point where sentiment significantly deteriorates.
Key requirements:
Sequential or graph-based thread modelling
Sentiment drift detection
Early-warning toxicity signals
Data sourced from platforms such as Reddit, X, or YouTube
Interpretable output showing which reply contributed to the sentiment shift
Prize Pool
Total Prize Pool: ₹10,000
Problem Statement
Prize
PS-01 — Tone Intelligence in Conversations
₹5,000
PS-02 — Voice as a Mood Signal
₹3,000
PS-03 — Sentiment Drift in Social Threads
₹2,000
Total
₹10,000
Schedule
Date & Time
Event
October 15, 2026
Registration opens on Unstop
November 1, 2026
Problem statements published
November 2, 2026 — 9:00 PM IST
Hackathon begins
November 3, 2026 — 9:00 AM IST
Halfway mark and optional mentor Q&A
November 4, 2026 — 3:00 AM IST
6 hours remaining
November 4, 2026 — 9:00 AM IST
Submissions close
November 4, 2026 — 12:00 PM IST
Judging begins
November 5, 2026
Winners announced
Hackathon Duration: 36 hoursFormat: Fully onlineRegistration: Free
What CEREBRO Is About:
CEREBRO is built around the intersection of AI and human emotion. Participants are encouraged to go beyond basic sentiment classification and explore how machines can understand changing emotional states across conversations, speech, and social interactions.
Participants are free to choose their own programming languages, frameworks, models, datasets, and development approaches.
The focus is on building a real, functional system rather than simply wrapping an existing AI API with a user interface.
Who Can Participate?
CEREBRO is open to students, developers, researchers, AI/ML enthusiasts, and anyone interested in building technology around emotion and sentiment intelligence.
Teams can participate in the hackathon and work remotely from anywhere.
Organized by
Department of Computer Science & Engineering
Institute of Engineering & Management (IEM), Kolkata
CEREBRO — Neural Sentiment HackathonNovember 2–4, 2026 · 36 Hours · Online