利用开源框架构建基于深度神经网络的短文本分类器  

Building Short Text Classification with Open Source Libraries Based on Deep Neural Network

作　　者：凡迪 付玉贞[1] 侯彤[1] 

机构地区：[1]四川省图书馆,成都610015

出　　处：《四川图书馆学报》2018年第1期23-25,共3页Journal of The Library Science Society of Sichuan

摘　　要：讨论了如何利用免费的开源代码库,快速构建基于深度神经网络的短文本分类器,可依照中文图书标题,将图书分类为文科和理工科两类。结果显示,文本分类器的准确率在测试集上的表现达91.9%。如果对开源社区的力量加以有效利用,能使图书馆行业充分受益。Open source libraries are freely available online. Based on these libraries,we were able to quickly build a binary book classifier with deep neural network. The classifier will tell if a book is on liberal arts subjects or science/technology subjects. We finally get an accuracy of 91. 9% on our testing set with the classifier. We believe that the library industry will definitely benefit,if librarians can make good use of the power of open source communities.

关 键 词：文本分类 深度神经网络 开源库 机器学习 情报学 

分 类 号：G250[文化科学—图书馆学]