Abstract
In this paper, we propose a method to extract indicated objects corresponding to demonstrative words for remote captioning. We extract indicated objects based on a trajectory of the edge position of a pointing stick and select the objects when a lecturer indicates the objects with demonstrative words.