BiEAF: An Bidirectional Enhanced Attention Flow Model for Question Answering Task

Question answering is an crucial module to construct an intelligent chatbot. Traditional question answering system utilizes various of deep learning models to answer the queries requested by users. While BiDAF modelled the interaction between context and query for the first time, which is a milestone attempt in this domain. However, BiDAF only takes the correlation across the sentences into account, but the association within the sentence is not considered. In this paper, based on BiDAF, we present a novel method to both utilize the inter-and-intra sentence interaction by deploying the proposed enhanced attention-flow layer. The experimental results on SQUAD dataset show that our method outperforms the baseline models in terms of both EM and F1 evaluation metrics, which proves our model effectiveness.