Joystick Text Entry with Word Prediction for People with Motor Impairments
Young Chol Song · 2010 · Proceedings of the 12th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2010) · doi:10.1145/1878803.1878892
Summary
This short paper presents a joystick-based text entry system enhanced with word completion and next word prediction, designed for people with motor impairments who cannot use conventional keyboards and mice. Existing joystick text entry methods — typically on-screen selection keyboards where users navigate a cursor across a character grid — require multiple joystick movements per character, resulting in slow input averaging about 6 words per minute. The system divides the display into three areas: a fixed character selection grid (5x6) showing the alphabet and minimal punctuation, a dynamic word prediction grid showing the fifteen most probable completions or next-word predictions, and a text area displaying entered text. Word prediction is powered by a trigram language model trained on a subset of the British National Corpus using a 40,000-word dictionary. Users navigate using a gamepad with two thumbsticks and eight buttons, moving a highlighted cursor through the grid. Two layout ordering strategies are supported: alphabetical (left to right, top to bottom) and probability-based (highest probability items centered in the grid, decreasing outward).
Key findings
Evaluation with four participants using four layout configurations showed that adding word prediction significantly improved performance. The baseline selection keyboard without prediction (Layout 1) achieved 5.82 words per minute, while the best-performing configuration — probability-ordered characters with probability-ordered predictions (Layout 4) — reached 8.06 wpm, a 30% speed improvement. Layout 4 also reduced joystick movements and button presses by 61.9% compared to the no-prediction baseline. Even alphabetical characters with probability-ordered predictions (Layout 3) achieved 50.8% fewer movements. Notably, these gains were achieved by first-time users with less than 15 minutes of practice, suggesting that performance would improve further with experience. Three of the four participants were first-time users of selection keyboard text entry. The results demonstrate that leveraging natural language structure through word prediction can meaningfully improve text entry speed without increasing the complexity of joystick operation.
Relevance
This research addresses a fundamental accessibility challenge: text entry speed for people who rely on alternative input devices. For motor-impaired users, every unnecessary joystick movement represents physical effort and fatigue, making the 50-62% reduction in movements particularly significant beyond just speed gains. The core principle — using language prediction to reduce physical input demands — has since become ubiquitous in mainstream technology (smartphone keyboards, voice assistants) but remains especially critical for assistive technology users. For accessibility practitioners designing text entry interfaces, the key takeaway is that probability-ordered layouts outperform alphabetical layouts when combined with word prediction, because they minimize the average distance the cursor must travel. The study also highlights an important gap: the evaluation used able-bodied participants, and the authors note that people with motor impairments using word prediction on standard keyboards sometimes show slower speeds due to higher cognitive load — suggesting that the cognitive cost of prediction must be weighed against its physical benefits.
Tags: motor impairment · text entry · word prediction · joystick · assistive technology · selection keyboard