A minutes taking system includes a change detection portion that detects a change of a displayed image for a description object out of displayed images that are a plurality of stream images serving as information displayed in a conference, a separation detection portion that detects a separation temporally separates a content of the displayed image; and an extraction portion that extracts a recorded time segment to be recorded as minutes information composing a minutes, on the basis of a change time when the change detection portion detects the change of the displayed image for the description object and a separation time when the separation detection portion detects the separation according to the plurality of streams.