How to find the right voice commands? Semantic priming in task representations

Abstract

A common way of interacting with conversational agents is through voice commands. Finding appropriate voice commands can be a challenge for developers and often require empirical methods where participants come up with voice commands for specific tasks. Textual and visual task representations were compared in an online study to assess the influence of semantic priming on spontaneous, user-generated voice commands. Our results indicate that visual representations of tasks lead to a different distribution of used phrases, compared to the textual representation, within the target population. Furthermore, the results suggest that text stimuli can be used to influence user utterances.