Abstract
This paper presents the Attention to Attention (A2A) reasoning mechanism to address the challenging task of movie question answering (MQA). By focusing on the various aspects of attention cues, we establish the technique of attention propagation to uncover latent but useful information to the underlying QA task. In addition, the proposed A2A reasoning seamlessly leads to effective fusion of different representation modalities about the data, and also can be conveniently constructed with popular neural network architectures. To tackle the out-of-vocabulary issue caused by the diverse language usages in nowadays movies, we adopt the GloVe mapping as a teacher model and establish a new and flexible word embedding based on character n-grams learning. Our method is evaluated on the MovieQA benchmark dataset and achieves the state-of-the-art accuracy for the “Video+Subtitles” entry.