1. Newly-Published Paper Recommendation With a Joint Multi-Relational Model
- Author
-
Wenjuan Ren, Dora. D. Liu, Qian Li, and Usman Naseem
- Subjects
Paper Recommendation ,tensor factorization ,relation model ,cold start ,personalization ,user behavior ,Electrical engineering. Electronics. Nuclear engineering ,TK1-9971 - Abstract
Reading newly-published papers in time is important for researchers since these papers provide the latest research findings. However, it is challenging to retrieve newly-published papers through common query-based search engines because the papers lacking sufficient citations and links are usually ranked too low in the search list. To this end, we design a time-aware joint model to infer users’ preference for the newly-published papers with the help of subsidiary relations of social and article linkages. The temporal preference of researchers for articles is jointly modeled with social and article relations by a group of matrices sharing common dimensions of researchers and articles. A joint multi-relational factorization algorithm is devised to approximate the latent factor matrices along with a temporal recommendation algorithm to predict the personalized new referential papers based on the factor matrices. The experimental results on real-world datasets show that the proposed model outperforms the state-of-the-art methods.
- Published
- 2022
- Full Text
- View/download PDF