Attention-based neural network for short-text question answering

Abstract

Question answering (QA) has been a popular topic in information retrieval tasks. Several studies rely on classifiers with a large number of handcrafted syntactic and semantic features and various external resources, such as WordNet, which is an English dictionary based on cognitive linguistics. Deep learning approaches have recently gained advanced performance in QA. However, these approaches have to be combined with additional features, such as word overlap. In this work, the factoid query answer retrieval task is introduced; moreover, the effective solving of this task under a deep learning framework is investigated. An attention-based convolutional neural network model is proposed to obtain word- and phrase-level interactive information and generate correct probability to re-rank candidate answers. The performance of the proposed model is compared with other models using the popular benchmark text retrieval conference QA data. Results show that the proposed model can obtain a significant performance improvement.