This dissertation presents a versatile design for text entry and control called EdgeWrite. EdgeWrite was designed to provide accessible text entry on a variety of platforms to people with motor impairments and to able-bodied users of small devices.

The EdgeWrite design includes a square input area, four corner sensors, cornersequence recognition, physical edges, goal crossing, and unistroke segmentation. Advanced EdgeWrite features include continuous recognition feedback, non-recognition retry, slip detection, and word-level stroking — concepts not before realized in a text entry method. The EdgeWrite alphabet, which is the same for all EdgeWrite versions, was designed for maximum guessability and learnability through participatory design. A key result is that EdgeWrite is about as guessable as Palm OS Graffiti, a method lauded for its immediate usability.

Multiple versions of EdgeWrite were built and user-tested, including Stylus EdgeWrite for the Palm OS, Joystick EdgeWrite for game controllers and power wheelchairs, Touchpad EdgeWrite, Trackball EdgeWrite, Isometric Joystick EdgeWrite for mobile phones, and EdgeWrite on four keys or sensors. For each version, empirical results from formal user studies were obtained. Of particular importance for users with motor impairments are the results for Stylus and Trackball EdgeWrite, which show marked improvements over existing techniques. In addition, five versions of EdgeWrite built by other researchers further highlight EdgeWrite’s versatility.

As part of EdgeWrite’s ongoing evaluations, a new character-level error analysis for text entry input streams was developed. This error analysis and the algorithms that automate it constitute a methodological and theoretical contribution to the field of text entry evaluation and measurement.

The thesis is: A versatile design for text entry and control called “EdgeWrite”, which uses physical edges, goal crossing, and a minimized need for sensing, is effective on handhelds and desktops for people with motor and situational impairments.

Full Thesis:
Download Jacob O. Wobbrock’s Full Thesis

Thesis Advisor:
Brad A. Myers

Award Date:
July 1, 2006

Carnegie Mellon University
Pittsburgh, Pennsylvania, USA

Author Contact: