KeyGlasses: Semi-transparent Keys on Soft Keyboard
Mathieu Raynal · 2014 · Proceedings of the 16th International ACM SIGACCESS Conference on Computers & Accessibility (ASSETS) · doi:10.1145/2661334.2661427
Summary
This paper presents KeyGlass, a text entry system designed to reduce pointer travel distance on soft (on-screen) keyboards by dynamically adding predicted character keys near the most recently typed character. Standard soft keyboards use QWERTY or AZERTY layouts familiar from physical keyboards, but these layouts are optimised for two-handed touch typing, not single-pointer input. When using a single pointer (mouse, head tracker, eye gaze, or single switch), users must travel long distances across the keyboard to reach frequently paired characters, leading to slower entry speeds, increased motor fatigue, and eye strain. Two prior approaches to this problem — optimised keyboard layouts (like Fitaly or OPTI) that rearrange keys by frequency, and word prediction lists — each have drawbacks: optimised layouts require learning an unfamiliar arrangement, and word prediction lists are positioned away from the keyboard, still requiring large pointer movements. KeyGlass bridges these approaches by displaying semi-transparent circular keys ("KeyGlasses") overlaid near the last entered character, showing the most probable next characters based on a prediction algorithm using lexicographic trees and bigram frequencies. The system is modular with three components: a fixed keyboard layout (configurable via XML), a prediction system that ranks probable next characters, and a display manager that positions KeyGlasses above the fixed keys. When a user types a character on either a fixed key or a KeyGlass, the system recursively generates new KeyGlasses for the next most likely characters. The semi-transparency ensures users maintain an overview of the full keyboard underneath, and KeyGlasses never overlap existing keys. The system uses the IVY software bus and is compatible with the E-ASSISTE assistive technology platform.
Key findings
The KeyGlass system demonstrates that character-level prediction can be integrated directly into the keyboard surface rather than presented as a separate word list, reducing the need for large pointer movements. By placing predicted characters adjacent to the last typed key, the system minimises traversal distance while preserving the familiar QWERTY/AZERTY layout as a fallback — users can always reach any character through the fixed keys if the prediction is incorrect. The recursive operation (each typed character, whether from a fixed key or KeyGlass, generates new predictions) means the system continuously adapts to minimise movement. The semi-transparent visual design is inspired by the "toolglass" concept from SIGGRAPH research, allowing predicted keys to augment the keyboard without obscuring it. The modular architecture enables easy substitution of prediction algorithms, keyboard layouts, or display strategies without rebuilding the entire system. The paper does not include user evaluation results, presenting the system as a design and technical contribution.
Relevance
KeyGlass addresses a significant accessibility challenge: efficient text entry for users who interact via a single pointer, which includes people using head-controlled mice, eye-gaze systems, mouth sticks, or single-switch scanning. For these users, every pixel of pointer travel costs time and physical effort, making keyboard layout optimisation directly relevant to productivity and fatigue management. The approach of bringing predicted characters to the pointer rather than requiring the pointer to travel to distant keys inverts the usual text entry paradigm and could benefit any single-pointer input scenario. For accessibility practitioners designing text entry interfaces, KeyGlass illustrates an important principle: the most accessible solution preserves the user's existing knowledge (familiar keyboard layout) while reducing physical demand (shorter pointer distances). The system's compatibility with the E-ASSISTE platform and modular design suggest it was intended for real-world assistive technology deployment, though the lack of user evaluation limits conclusions about its practical effectiveness compared to alternatives like word prediction or optimised layouts.
Tags: text entry · soft keyboard · character prediction · motor accessibility · input methods · assistive technology