Modelling Learning of New Keyboard Layouts
summary
Computational model predicts how people move from visual-search to memory-guided key selection, combining vision + visual short‑term memory + long‑term memory with a relearning controller that downweights old associations after layout changes. Fitted mostly from literature, tested against a 5‑day user study, it predicts search-time drops (e.g., ~1.2s to ~0.6s on QWERTY) and simulates swaps, Dvorak transitions and soft‑keyboard variants. Model and code are available on the project website or by email.
AI-written summary · about
chapters (9)
- 0:00Introduction and motivation
- 1:58Learning and relearning layouts
- 3:28Visual search patterns example
- 6:37Model goals and requirements
- 8:04Model components overview
- 11:44Testing and validation
- 13:56Layout comparisons
- 15:22Summary and applications
- 16:08Q&A session
related talks
13mUniKey: Enabling Surface-Based Typing with Commodity Smartwatches via Cross-Modal Learning
21mThe Impact of Word, Multiple Word, and Sentence Input on Virtual Keyboard Decoding Performance
03mInterview with Ben Medlock
37mA self learning word prediction module for handling multilingual input
05mExploring Smartphone Keyboard Interactions for Experience Sampling Method driven Probe Generation
40mHuman Learning
27mSign Language Recognition Improvements Since the London Keynote
28mPreference Learning from Minimal Human Feedback for Interactive