Attention-based CNN Matching Net

In this paper, we introduce attention-based CNN matching net (ACM-Net), an end-to-end neural network for question answering. ACM-Net matches between the given passage, query and multiple answer choices, and then it extracts features from passage and choices based on query information. We also propose a two-staged CNN architecture and a query-based attention mechanism in our model. These two component can effectively find out the most important parts in passage according to the query. Finally, we extract features from those important parts and find out the most possible answer choice. We conduct this model on the MovieQA dataset using Plot Synopses only, and achieve 79.99% accuracy which is the state of the art on the dataset.