Speech recognition technology continues to improve, but users still experience significant difficulty using the software to create and edit documents. In fact, a recent study confirmed that users spent 66% of their time on correction activities and only 33% on dictation. Of particular interest is the fact that one third of the users' time was spent simply navigating from one location to another. In this article, Andrew Sears is a computer scientist with an interest in human-computer interaction in the areas of speech recognition, mobile computing, and accessibility; he is a Professor in the Information Systems Department of UMBC. Jinjuan Feng is interested in human-computer interaction; she is a PhD student in the Information Systems Department of UMBC. Kwesi Oseitutu is interested in human-computer interaction; he received an MS from the Information Systems Department of UMBC. Clare-Marie Karat is a social psychologist with an interest in human-computer interaction in the areas of speech recognition, privacy and security, and personalization; she is a Research Staff Member at IBM's T.J. Watson Research Center. Downloaded by [North West University] at 14:49 20 December 2014we investigate the efficacy of hands-free, speech-based navigation in the context of dictation-oriented activities. We provide detailed data regarding failure rates, reasons for failures, and the consequences of these failures. Our results confirm that direction-oriented navigation (e.g., Move up two lines) is less effective than target-oriented navigation (e.g. Select target). We identify the three most common reasons behind the failure of speech-based navigation commands: recognition errors, issuing of invalid commands, and pausing in the middle of issuing a command. We also document the consequences of failed speech-based navigation commands. As a result of this analysis, we identify changes that will reduce failure rates and lessen the consequences of some remaining failures. We also propose a more substantial set of changes to simplify direction-based navigation and enhance the target-based navigation. The efficacy of this final set of recommendations must be evaluated through future empirical studies.

Speech recognition is an important technology that is becoming increasingly effective for dictationoriented activities. While recognition accuracy has increased dramatically in recent years, recent studies confirm that traditional computer users are still faster using a keyboard and mouse and spend more time correcting errors than dictating. Further, as these users become more experienced they frequently adopt multimodal strategies that require the keyboard and mouse when correcting errors. While speech recognition can be a convenient alternative for traditional computer users, it can be a powerful tool for individuals with physical disabilities that limit their ability to use a keyboard and mouse. However, research into the performance, satisfaction, and usage patterns of individuals with physical disabilities has not been reported. In this article, we report on a study that provides initial insights into the efficacy of existing speech recognition systems with respect to individuals with physical disabilities. Our results confirm that productivity does not differ between traditional users and those with physical disabilities. In contrast, numerous differences were observed when users rated their satisfaction with the system and when usage patterns were analyzed.