Reconstruct and Represent Video Contents for Captioning via Reinforcement Learning | IEEE Journals & Magazine | IEEE Xplore