图书情报工作 ›› 2017, Vol. 61 ›› Issue (23): 129-137.DOI: 10.13266/j.issn.0252-3116.2017.23.016

• 知识组织 • 上一篇    下一篇

基于Spark的高校图书馆文献推荐方案及实证研究

何胜1,2, 熊太纯3, 柳益君1,2, 叶飞跃1,2, 潘瑜1,2   

  1. 1. 江苏理工学院计算机工程学院 常州 213001;
    2. 常州市云计算与智能信息处理重点实验室 常州 213001;
    3. 江苏理工学院图书馆 常州 213001
  • 收稿日期:2017-06-12 修回日期:2017-09-18 出版日期:2017-12-05 发布日期:2017-12-05
  • 作者简介:何胜(ORCID:0000-0001-6762-8271),副教授,博士,E-mail:hs@jsut.edu.cn;熊太纯(ORCID:0000-0002-8623-4793),研究馆员;柳益君(ORCID:0000-0002-5381-6084),副教授;叶飞跃(ORCID:0000-0002-6068-7567),教授;潘瑜(ORCID:0000-0002-5817-8675),教授。
  • 基金资助:
    本文系国家社会科学基金一般项目"基于大规模网络分析方法和内存计算技术的高校图书馆大数据应用模式与实证研究"(项目编号:15BTQ016)研究成果之一。

A Spark-based Scheme of University Library Literature Recommendation and Its Empirical Study

He Sheng1,2, Xiong Taichun3, Liu Yijun1,2, Ye Feiyue1,2, Pan Yu1,2   

  1. 1. School of Computer Engineering, Jiangsu University of Technology, Changzhou 213001;
    2. Key Laboratory of Cloud Computing & Intelligent Information Processing of Changzhou City, Changzhou 213001;
    3. Jiangsu University of Technology Library, Changzhou 213001
  • Received:2017-06-12 Revised:2017-09-18 Online:2017-12-05 Published:2017-12-05

摘要: [目的/意义]以高校图书馆馆藏图书数据库和各类论文数据库等海量的文献资源为基础,设计推荐方案并基于Spark技术开展实证研究,力图优化图书馆文献推荐效果和提高系统计算性能。[方法/过程]首先分析大数据背景下高校图书馆文献推荐的需求,接着针对存在的文献查找缺失、文献浏览迷航和文献分析低效的现状,提出一种以文献"混合关联"为主要内容的图书馆文献推荐方案及实现算法,并应用Spark内存计算技术设计实证案例,最后对实证结果进行讨论并与同类算法比较。[结果/结论]基于Spark的文献"混合关联"方案能有效满足用户需求,提高文献推荐性能和效率,促进当前图书馆大数据应用的落地。

关键词: 图书馆文献推荐, 混合关联, 大数据, 内存计算, Spark

Abstract: [Purpose/significance] In order to improve recommendation effects and the computation performance, a recommendation scheme is designed and its empirical study is realized in this paper based on the mass literature resource including bibliographic databases and various paper databases. [Method/process] Firstly, the paper analyzed the requirement of the literature recommendation of university libraries under the big data background. Then, it put forward the scheme and its algorithm implementation of library literature recommendation with the hybrid link strategy to solve the problems of the literature query deficit, the literature browse loss and the literature analysis inefficiency, and designed a case study based on in-memory computing technology of Spark. Finally, the paper discussed results of the experiment after comparison with similar algorithms. [Result/conclusion] The proposed scheme can meet the users' requirements efficiently and improve the performance and efficiency of literature recommendation and promote the application of big data in libraries at present.

Key words: library literature recommendation, hybrid link, big data, in-memory computing, Spark

中图分类号: