BYU

Abstract by Eric Brewer

Personal Infomation


Presenter's Name

Eric Brewer

Co-Presenters

Alisha Banskota

Degree Level

Masters

Co-Authors

Alisha Banskota
Yiu-Kai Ng

Abstract Infomation


Department

Computer Science

Faculty Advisor

Dennis Ng

Title

Personalized Scholarly Article Recommendations Based on the Recurrent Neural Network and Probabilistic Models

Abstract

Searching scholarly articles, i.e., research papers, on the Web is a challenge even for researchers who are familiar with the
searched topics and articles of a particular domain of interest, needless to say for users who are not familiar with the
areas of study. Existing research paper recommendation systems either rely on the content-based or collaborative-filtering
approach, or its hybrid model. These recommenders, however, are vulnerable to the cold-start problem, either on new users
or new scholarly articles. In this paper, we propose an innovative personalized scholarly article recommendation system
which suggests research papers of the same subject area using the recurrent neural network model and ranks closely related
research papers using the BM25 probabilistic model. Experimental results based on the titles and abstracts of research
papers extracted from ACM DL and IEEE Xplore digital library verify the merit of the proposed recommender, which
outperforms recently-developed research paper recommendation systems.