图书情报工作 ›› 2022, Vol. 66 ›› Issue (11): 110-120.DOI: 10.13266/j.issn.0252-3116.2022.11.012

• 情报研究 • 上一篇    下一篇

虚拟学术社区中融合用户动态兴趣与社交关系的学者推荐研究

顾佳云, 熊回香, 肖兵   

  1. 华中师范大学信息管理学院 武汉 430079
  • 收稿日期:2021-10-07 修回日期:2022-02-21 出版日期:2022-06-05 发布日期:2022-06-18
  • 通讯作者: 熊回香,教授,博士,博士生导师,通信作者,E-mail:hxxiong@mail.ccnu.edu.cn。
  • 作者简介:顾佳云,硕士研究生;肖兵,博士研究生。
  • 基金资助:
    本文系国家社会科学基金重大项目"新时代我国文献信息资源保障体系重构研究"(项目编号:19ZDA345)研究成果之一。

Research on Scholar Recommendation Integrating Users' Dynamic Interests and Social Relationships

Gu Jiayun, Xiong Huixiang, Xiao Bing   

  1. School of Information Management, Central China Normal University, Wuhan 430079
  • Received:2021-10-07 Revised:2022-02-21 Online:2022-06-05 Published:2022-06-18

摘要: [目的/意义] 考虑用户兴趣和社交关系两方面的动态变化,提出融合用户动态兴趣与社交关系的学者推荐模型。[方法/过程] 首先,利用不同学科的期刊文献作为分类语料,基于Labeled-LDA模型对学者所发博文进行学科领域判别。然后,依据KNN算法对博文进行学科分类,接着利用学科兴趣变化速率改进时间因子,计算得到学者动态兴趣相似度;根据学者间链接的数量关系计算学者的PageRank值,结合学者所发博文的时间价值计算得到全局信任度。在学者评论、推荐交互行为中引入时间权重计算学者交互信任度,综合全局信任度和交互信任度得到学者的动态社交信任度。最后,融合兴趣相似度与信任度进行学者推荐。[结果/结论] 虚拟学术社区中融合用户动态兴趣与社交关系的学者推荐模型从动态兴趣和动态社交关系两个视角出发,能够有效提高学者推荐的质量。

关键词: 虚拟学术社区, 动态兴趣, 社交关系, 学者推荐, Labeled-LDA主题模型

Abstract: [Purpose/Significance] Considering the dynamic changes of users’ interests and social relationships, this paper proposes a scholar recommendation model integrating users’ dynamic interests and social relationships. [Method/Process] Firstly, using the periodical literature of different disciplines as the classified corpus, the discipline domain of scholars’ blog posts was distinguished based on the labeled LDA model. Then KNN algorithm was used to classify blogs by discipline. At the same time, the change rate of subject interests was used to improve the time factor, and the dynamic interest similarity of scholars was calculated. The PageRank of scholars was calculated by using the quantitative relationship of links between scholars, and the global trust level was calculated by combining the PageRank and time value of blogs sent by scholars. Time weight was introduced into scholars’ comments and recommendation interaction behaviors to calculate scholars’ interactive trust level. The dynamic social trust level of scholars was obtained by integrating the global trust level and interactive trust level. Finally, the similarity of interest and trust were combined to recommend scholars. [Result/Conclusion] The scholar recommendation model integrating users’ dynamic interests and social relationships in the virtual academic community can effectively improve the quality of scholar recommendation from the perspectives of dynamic interests and dynamic social relationships.

Key words: virtual academic community, dynamic interests, social connections, scholar recommendation, Labeled-LDA topic model

中图分类号: