Large Vocabulary Shorthand Writing on Stylus Keyboard

Abstract: We present a novel text entry method for pen-based computers. We view the trace obtained by connecting the letter keys comprising a word on a stylus keyboard as a pattern. This pattern can be matched against a user-s pen trace, invariant of scale and translation. Hence the patterns become an efficient form of shorthand gestures, allowing users to use eyes-free openloop motor actions to perform the gestures. This can result in higher text entry speed than optimized stylus keyboards, the fastest known text entry technique for pen-computers as of today. The approach supports a gradual and seamless skill transition from novices tracing the letter keys to experts articulating the shorthand gestures. Hence the ratio between the learning effort and efficiency in using the system can be said to be optimized at any given point in time in the user-s experience with the technique. This thesis describes the rationale, architecture and lgorithms behind a stylus keyboard augmented with a high-capacity gesture recognition engine. We also report results from an Expanding Rehearsal Interval (ERI) experiment which indicates that users can acquire about 15 shorthand gestures per 45 minute training session. Empirical expert speed estimates of the technique indicate text entry speeds much higher than any prior known pen-based text entry system for mobile computers.