A pattern language for error management in voice user interfaces

Speech is not recognized with an accuracy of 100%. Even humans are not able to do that. There will always be some uncertainty in the recognized input, requiring strategies to cope. This is different from the experience with graphical user interfaces, where keyboard and mouse input are recognized without any doubts. Speech recognition and other errors occur frequently and reduce both the usefulness of applications and user satisfaction. This turns error handling into a crucial aspect of speech applications. Successful error handling methods can make even applications with poor recognition accuracy more successful. In [Sagawa et al. 2004] the authors show that the task completion rate increased from 86.4% to 93.4% and the average number of turns reduced by three after a better error handling method had been installed. On the other hand, poorly constructed error handling may bring unwanted complexity to the system and cause new errors and annoyances.