View article

Using speech input to augment the remote control can be an alternative interaction technique for interactive television. However, little is known about how to design such a system that is suitable for the home environment. In this paper we explore possible error resolution strategies in the case of speech recognition errors in a TV-setting. From previous research two techniques have been identified: repetition of input by the user and choice of interpretation. From these two techniques, four alternative strategies, suitable for a TV-environment, were designed and tested with a Wizard-of-Oz method: one with repetition and three based on choosing the best alternative from an n-best list given in audio and visual mode, alone or combined. The results show that displaying an n-best list gives the most efficient interaction. Redundant audio feedback does not influence the performance.