Multimodal Learning for Accurate Visual Question Answering: An Attention-Based Approach