While the growth of the smart speaker market is accelerating with the development of voice interaction technology, users' satisfaction and continuity of use are low. This study explored the need for command cancellation and speech interruption through interviews and observations based on repeated call word (wake-up call) and conversational structural problems of smart speakers, and attempted to improve error situations by deriving a re-response activation structure model. Through the experiment, the error situation of each use case was found, and through the in-depth interview, the problems of the conversation structure of the smart speaker and the needs of the user were embodied to suggest improvements for smooth voice interaction. The study is meaningful in that it analyzed the dialogue structure of actual smart speakers as a