基于卷积神经网络的跨领域语义信息检索研究  

CROSS-DOMAIN SEMANTIC INFORMATION RETRIEVAL BASED ON CONVOLUTIONAL NEURAL NETWORK

作　　者：谢先章 王兆凯[1] 李亚星[1] 冯旭鹏 刘利军[1] 黄青松[1,3] Xie Xianzhang1,Wang Zhaokai1,Li Yaxing1,Feng Xupeng3,Liu Lijun1,Huang Qingsong1,2(1.Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, Yunan,China;2.Yunnan Key Laboratory of Computer Technology Applications, Kunming 650500, Yunan,China;3.Educational Technology and Network Center, Kunming University of Science and Technology, Kunming 650500, Yunan, Chin)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500 [2]昆明理工大学教育技术与网络中心,云南昆明650500 [3]云南省计算机技术应用重点实验室,云南昆明650500

出　　处：《计算机应用与软件》2018年第8期73-78,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(81360230;81560296)

摘　　要：随着社会信息化的发展,信息检索成为了影响人们日常生活和生产工作的重要技术,同时人们对信息检索技术的要求也越来越高[1]。为了改进传统检索方式,基于卷积神经网络CNNs(Convolutional Neural Networks)提出一种语义信息检索模型,对句子卷积特征和词聚特征进行映射训练,达到搜索近似语句的目的。实验表明基于卷积神经网络的近似语义检索模型相比传统检索模型有助于提高信息检索的质量。With the development of information society, information retrieval has been made an important technique which influences our daily life and work. At the same time, people have increasingly high requirements for the technique of information retrieval. In order to improve the methods of traditional retrieval, this paper presented a semantic information retrieval model based on the convolutional neural network which made mapping training in sentence convolution dimension and word clustering dimension to achieve the purpose of searching approximate statement. Experiments show that approximate semantic retrieval model based on convolution neural network is useful to improve the quality of information retrieval compared with the traditional retrieval model.

关 键 词：卷积神经网络 信息检索 词向量 SVM 快速聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]