Heterogeneous hypergraph embedding for document recommendation

详细信息    查看全文

• 作者：Yu Zhu^a ; ^{zy11120085@gmail.com" class="auth_mail" title="E-mail the corresponding author}Author Vitae ; Ziyu Guan^b ; ^{welbyhebei@gmail.com" class="auth_mail" title="E-mail the corresponding author}Author Vitae ; Shulong Tan^c ; ^{laos1984@gmail.com" class="auth_mail" title="E-mail the corresponding author}Author Vitae ; Haifeng Liu^d ; ^{haifengliu@zju.edu.cn" class="auth_mail" title="E-mail the corresponding author}Author Vitae ; Deng Cai^a ; ^{dengcai@cad.zju.edu.cn" class="auth_mail" title="E-mail the corresponding author}Author Vitae ; Xiaofei He^a ; ^{xiaofeihe@gmail.com" class="auth_mail" title="E-mail the corresponding author}Author Vitae
• 关键词：Recommender systems ; Hypergraph ; Graph-based learning
• 刊名：Neurocomputing
• 出版年：2016
• 出版时间：5 December 2016
• 年：2016
• 卷：216
• 期：Complete
• 页码：150-162
• 全文大小：2068 K

文摘

Nowadays, more and more users are using online tagging services to organize their resources, e.g. Web bookmarks and bibliographies. Tags not only facilitate organization and retrieval of resources, but also provide valuable semantic descriptions for both resources and users’ interests. This work is focused on document recommendation using tagging data. Previous works either model the 3-order relation <$<$user, tag, document  >$>$ in tagging data by an ordinary graph or model different types of relations by a homogeneous hypergraph. The former scheme would lead to serious information loss, and the latter one fails to discern the influence of different types of relations. In this paper, we propose a heterogeneous hypergraph model which fully exploits high-order relational information in tagging data and, meanwhile, customizes the influence of different types of relations. A novel heterogeneous hypergraph embedding framework is developed for document recommendation. The framework is general and can incorporate various relations among users, tags and resources. Experimental results on two real-world datasets show the superiority of the proposed method over traditional methods.