A Hybrid of Deep Sentence Representation and Local Feature Representation Model for Question Answer Selection

Answer selection is a one of the critical tasks in natural lan-guage processing area and it is helpful in many practical applications. To better tackle this problem, the first challenge is to effectively extract the sentence information. In this research, we propose an advanced Re-Read-CNN model which can learn a deep sentence representation and meanwhile combine the local feature representation. The experiment results on commonly used datasets have shown its effectiveness and potential for answer selection.