Emotion Typing
Detect emotions from typing patterns: speed, rhythm, pauses.
Duration:
4 weeks
Group size:
4-5 students
Outcome:
Get a fully functional machine learning based web application which predicts emotion from typing.
Prerequisites:
High GPU Computer or Laptop. No tablets/iPads
Fees:
Please contact us at team@create-ed.in
Week by week curriculum
Session 1
Understand keystroke dynamics, define target emotions, choose a dataset or plan custom data collection, and set up your project environment.
Session 2
Build a Python/JS keystroke logger to capture keydown/keyup timings and gather labeled typing samples across different emotional states.
Session 3
Transform raw keystroke events into numerical features like dwell time, flight time, typing speed, and variability, and clean the dataset for modeling.
Session 4
Train classical models (RandomForest, SVM, Logistic Regression), evaluate with accuracy/F1, and save the best-performing baseline.
Session 5
Build and train LSTM/GRU or 1D CNN models for sequential keystroke data, compare results against baseline ML models, and export the best model.
Session 6
Create a live prediction system that logs keystrokes, computes features on the fly, smooths outputs, and streams continuous emotion predictions.
Session 7
Develop a Gradio interface with a typing box, real-time keystroke capture, and UI elements showing emotion predictions, probabilities, and simple charts.
Session 8
Deploy the app on HuggingFace Spaces, refine styling, improve responsiveness, and run user tests to ensure accurate predictions and a clean experience.