Video Captioning by Adversarial LSTM