In this paper, we are exploring a human-robot in- terface allowing non-expert users to easily and intu- itively teach new words to a robot. Many challenges may be addressed before achieving such behavior as joint attention, naming, categorization and search- ing. Instead of using direct interactions (such as gesture and voice recognition) which are not robust enough in unconstrained environments, we use here a handheld device as a mediator between the user and the robot. The device allows us, among other things, to display the robot's camera visual feed- back as well as interacting through user's gestures on the touchscreen. Thus, users are able to draw the robot's attention toward locations, to select a particular ob ject in the field of view of the robot, to name it and with our “active-searching” system to find it again later. An exploratory study has been carried out in order to get some early users opinions, which tend to show that users easily use the interface during robust interaction sessions.