Modelling children's natural interactions from a video corpus