Abstract | ||
---|---|---|
Rating prediction is a common task in recommendation systems that aims to predict a rating representing the opinion from a user to an item. In this paper, we propose a comment-based collaborative filtering (CCF) approach that captures correlations between hidden aspects in review comments and numeric ratings. The idea is motivated by the observation that the opinion of a user against an item is represented by different aspects discussed in review comments. In our approach, we first explores topic modeling to discover hidden aspects from review comments. Profiles are then created for users and items separately based on the discovered aspects. In the testing stage, we estimate the aspects of comments based on the profiles of users and items because the comments are not available when testing. Lastly, we build final systems by utilizing the profiles and traditional collaborative filtering methods. We evaluate the proposed approach on a real data set. The experimental results show that our prediction systems outperform several strong baseline systems. |
Year | DOI | Venue |
---|---|---|
2015 | 10.1007/978-3-319-21042-1_20 | WEB-AGE INFORMATION MANAGEMENT (WAIM 2015) |
Field | DocType | Volume |
Recommender system,Latent Dirichlet allocation,Collaborative filtering,Computer science,Mean absolute error,Artificial intelligence,Topic model,Random forest,Machine learning | Conference | 9098 |
ISSN | Citations | PageRank |
0302-9743 | 2 | 0.36 |
References | Authors | |
13 | 7 |
Name | Order | Citations | PageRank |
---|---|---|---|
Rong Zhang | 1 | 39 | 6.77 |
Yifan Gao | 2 | 2 | 0.70 |
Wenzhe Yu | 3 | 8 | 1.46 |
Pingfu Chao | 4 | 5 | 3.13 |
Xiaoyan Yang | 5 | 89 | 5.79 |
Ming Gao | 6 | 76 | 9.41 |
Aoying Zhou | 7 | 2632 | 238.85 |