A Question Answering System Based on Deep Learning

Abstract

The goal of the question answering system is to automatically answer questions posed by humans being in the natural language text. Recently, the restricted domain question answering system has become a research hotspot. In this paper, we examine how to efficiently use deep learning method to improve the performance of the question answering system, and design a legal question answering system. Firstly, initial results are obtained by using the vector space model, and then the results are rendered through similarity between the answers. Finally, in order to optimize the system, the deep convolutional neural network is adopted to obtain a one-dimensional sentence vector, which is used to replace the keyword vector for the answer candidate. Experimental results show that the proposed method outperforms the traditional keyword vector method.