关键词: 问答匹配, 深度学习, 注意力机制, 循环神经网络, 卷积神经网络

Abstract: Intelligent Question Answering is an important part of Natural Language Processing task.The Chinese Question Answer matching based on traditional methods has the problem of high labor dependence, poor efficiency and low accuracy.Using deep learning method to learn the abstract semantic information in sentence has obvious advantages ,as well as attention mechanism based model can concern the critical semantic information of words in sentence,reducing the impact of some non-critical noise words on matching results.For question answer matching,This paper propose deep learning model based on recurrent neural network (RNN) and convolution neural network (CNN) with combining attention mechanism.It can effectively learn word order, semantics and key information characteristics in the sentence.Experimental results show that it still has good results and high accuracy of question answer matching on the condition that without manual additional rules to formulate the characteristic.

Key words: Question Answer Matching, deep learning, attention mechanism, recurrent neural network, convolution neural network