图书情报工作 ›› 2016, Vol. 60 ›› Issue (17): 130-135.DOI: 10.13266/j.issn.0252-3116.2016.17.019

• 知识组织 • 上一篇    下一篇

基于相对频次的标签相关性判断优化研究

林鑫1, 石宇2, 周知2   

  1. 1. 华中师范大学信息管理学院 武汉 430079;
    2. 武汉大学信息管理学院 武汉 430072
  • 收稿日期:2016-02-15 修回日期:2016-05-10 出版日期:2016-09-05 发布日期:2016-09-05
  • 作者简介:林鑫(ORCID:0000-0003-0318-8160),讲师,E-mail:wenqi0426@qq.com;石宇(ORCID:0000-0001-6741-5487),硕士研究生;周知(ORCID:0000-0002-5530-2968),博士研究生。

Study on Social Tags Relevance Judgment Optimization Based on Relative Frequency

Lin Xin1, Shi Yu2, Zhou Zhi2   

  1. 1. School of information management of Central China Normal University, Wuhan 430079;
    2. School of Information Management of Wuhan University, Wuhan 430072
  • Received:2016-02-15 Revised:2016-05-10 Online:2016-09-05 Published:2016-09-05

摘要:

[目的/意义] 针对笔者此前提出的标签相关性判断策略进行优化,以提升策略的召回率,从而更好地支持标签应用研究与实践。[方法/过程] 为提升策略的召回率,以标签与认知的基本关系为基础,提出一种基于相对频次的改进策略,并以社会化标注社区"豆瓣电影"的675 351位用户的标签数据为例进行实验,以验证策略的效果。[结果/结论] 结果显示,该策略使得标签相关性判断的效果得到了显著改善。其中,对于频次不小于5的标签,策略的召回率大幅提升,由79.63%升至89.36%;准确率虽有略微下滑,由93.33%降至92.02%,但仍保持在较高水平。

关键词: 社会化标签, 标签相关性, 相对频次

Abstract:

[Purpose/significance] To improve the recall and support the tag application research and practice, this paper optimizes the tactic proposed in our previous paper for social tags relevance judgment. [Method/process] To improve the recall, this paper took the relationship between social tags and cognition as the basis, and proposed an optimization algorithm based on relative frequency, and verified by experiments based on 675 351 Douban Movie users' social tagging data. [Result/conclusion] The results show that the effect of tag relevance judgment for tags whose frequencies are at least 5 has been improved significantly, for the recall rises sharply from 79.63% to 89.36%, and the accuracy rate falls slightly from 93.33% to 92.02%, but remains at a high level.

Key words: social tagging, tag relevance, relative frequency

中图分类号: