Embodiments of the present disclosure provide a method and apparatus for segmenting a video. The method may include: performing video shot segmentation on a to-be-segmented video to obtain an initial video clip set; selecting at least one video clip from the video clip set; determining, for a video clip in the at least one video clip, a semantic similarity between two video clips adjacent to the video clip; and combining two video clips corresponding to a semantic similarity greater than a preset first similarity threshold value in the determined semantic similarity and a video clip between the two video clips to obtain a final video clip set.