Expecting certain behaviors from users is essential when programming the actions of communication robots. However, an ef̌ cient way to bring forth this expectation has not been realized yet. To tackle this problem, this paper presents a novel sensor data analysis system called RSV (Robot Sensor Viewer). RSV enables the discovery of user behaviors by comparing various enormous data samples. RSV associates user behaviors with the sensor data and it enables incorporation of detected behaviors into the robots' action programs. To evaluate RSV, we conducted a communication experiment, obtained the data samples of user behaviors and analyzed them by using RSV. The results from the experiment showed that RSV could detect the features of user behaviors which were not considered in the current action programs on communication robot Robovie.