(Georgia Institute of Technology, 2000)
Mankoff, Jennifer C.; Hudson, Scott E.; Abowd, Gregory D.
Because of its promise of natural interaction, recognition is coming into its own as a mainstream technology for use with computers. Both commercial and research applications are beginning to use it extensively. However the errors made by recognizers can be quite costly, and this is increasingly becoming a focus for researchers. We present a survey of existing error correction techniques in the user interface. These mediation techniques most commonly fall into one of two strategies, repetition and choice. Based on the needs uncovered by this survey, we have developed OOPS, a toolkit that supports resolution of input ambiguity through mediation. This paper describes four new interaction techniques built using OOPS, and the toolkit mechanisms required to build them. These interaction techniques each address problems not directly handled by standard approaches to mediation, and can all be re-used in a variety of settings.
(Georgia Institute of Technology, 1999)
Mankoff, Jennifer C.; Abowd, Gregory D.; Hudson, Scott E.
Despite significant advances in recognition technologies in areas such as speech and gesture recognition, our experience tells us that recognition errors and uncertainty are unlikely to disappear. For the foreseeable future, use of recognition based systems will introduce uncertainty into the input process. If interactive systems are going to work robustly with recognition-based input, it will be necessary to consider uncertainty as a normal part of input handling rather than considering it to be an anomaly or an exceptional condition. This paper considers techniques for explicit treatment of input uncertainty in user interfaces. In particular, it considers a general class of techniques for the display of, and interaction with, multiple alternatives generated by recognition technologies. Augmentation of the typical event-handling infrastructure is discussed, as well as an application interface infrastructure which attempts to minimize the impact of uncertainty on the application. A prototype system that embodies this infrastructure is also considered.