情报科学

• • 上一篇    下一篇

基于用户相似度与信任度的虚拟学术社区中学者推荐研究

  

  1. 华中师范大学信息管理学院

Research on scholar recommendation in virtual academic community based on user similarity and trust

  1. School of Information Management, Central China Normal University

摘要:

目的/意义研究从用户节点和网络全局两个视角出发,基于用户相似度与信任度对虚拟学术社区中学者进行推荐,提高学者推荐的质量。方法/过程首先,利用LDA主题模型挖掘学者发表的博文主题,计算博文相似度;通过学者共同好友比例计算好友相似度;然后将博文相似度和好友相似度融合计算用户相似度;最后,融合用户相似度和信任度进行学者推荐。结果/结论提出虚拟学术社区中基于用户相似度与信任度的学者推荐方法,综合利用用户节点和网络全局信息,为虚拟学术社区用户进行学者推荐创新/局限从用户节点和网络全局两个角度进行学者信息融合,有效提高了虚拟学术社区中学者推荐的质量。局限在于本文主要考虑的是学者在网络全局中的信任度,用户节点间的交互信任关系还有待进一步研究。

关键词:

虚拟学术社区, 学者推荐, LDA主题模型, 信任度, 用户相似度

Abstract:

Purpose/significance】From the perspective of user nodes and the overall network, the study makes recommendations to scholars in virtual academic communities based on the user similarity and trust, which improves the quality of scholar recommendations. 【Method/process】Firstly, LDA topic model is used to mine the blog topics published by scholars and calculate the similarity of blog posts. The similarity degree of friends is calculated by the ratio of common friends of scholars. Then the similarity of the blog and the similarity of the friends are combined to calculate the user similarity. Finally, we integrate user similarity and trust to implement scholar recommendation. 【Result/conclusion】This paper proposes a scholar recommendation method based on user similarity and trust in virtual academic community, which makes comprehensive use of user nodes and network global information to recommend scholars for users in virtual academic community. 【Innovation/limitation】The fusion of scholars' information from two perspectives of user nodes and network overall improves the quality of scholars recommended in the virtual academic community. The limitation is that this paper mainly considers the trust degree of scholars in the whole network, and the mutual trust relationship between user nodes needs to be further studied.

Key words:

Virtual Academic Community, Scholar Recommendation, LDA topic model, Trust, User similarity