面向社区问答匹配的混合神经网络模型

Hybrid Neural Network Model for Community Question/Answer Matching

问答匹配是社区问答的一项重要且具有挑战性的任务.本文提出了一种面向社区问答匹配的混合神经网络模型.针对问答对序列,提出了融合卷积神经网络(CNN)与双向长短期记忆网络(Bi-LSTM)的混合模型,学习问答对的语义信息及问答对序列的上下文相关性信息;针对用户的历史回答,提出基于多维度注意力机制的用户-问题建模方法,学习用户与问题之间的相关性信息.在SemEval-2015CQA数据集上的实验结果表明,与现有的社区问答匹配算法相比,本文算法能够有效提高社区问答匹配精度.

Question and answer matching is an important and challenging task of community question answering(CQA).In this paper,we propose a hybrid neural network model for community question/answer matching.For question and answer pairs,a hybrid model of fusion convolution neural network(CNN)and bi-directional long short-term memory network(Bi-LSTM)is proposed to learn the semantic information of question and answer pair and the contextual relevance information of the question and answer sequence.According to the historical answer of the user,a user-question modeling method based on multi-dimensional attention mechanism is proposed to learn the correlation information between the user and the question.The experimental results on SemEval-2015 CQA dataset show that the proposed algorithm can effectively improve the accuracy of CQA matching compared with the existing CQA matching algorithms.