Learning to utilize auxiliary reviews for recommendation

Review-based recommender systems represent users and items with reviews associated with them. As such, the recommender systems are highly dependent on the number of reviews, which is usually few in number. Thus, they produce inaccurate recommendations to users who rarely wrote reviews. An approach t...

Full description

Saved in:
Bibliographic Details
Published in:Information sciences 2021-02, Vol.545, p.595-607
Main Authors: Hyun, Dongmin, Park, Chanyoung, Cho, Junsu, Yu, Hwanjo
Format: Article
Language:English
Subjects:
Attention mechanism
Deep learning
Recommender system
Review
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Review-based recommender systems represent users and items with reviews associated with them. As such, the recommender systems are highly dependent on the number of reviews, which is usually few in number. Thus, they produce inaccurate recommendations to users who rarely wrote reviews. An approach to generate better recommendations for the cold-start users is to augment the scarce reviews using other reviews, which are called auxiliary reviews. In this work, we address two research questions when leveraging auxiliary reviews: 1) How to find important auxiliary reviews? 2) How to combine the auxiliary reviews with their original reviews?, and propose a method that learns to utilize auxiliary reviews (LUAR) to tackle these questions. LUAR learns to automatically focus on important auxiliary reviews from data via neural attention mechanism. For the review combination issue, a self-attentive module in LUAR combines auxiliary reviews and original reviews considering their level of contribution. The module dynamically computes the level of contribution of each review based on its relative importance compared to others. Experimental results show that LUAR outperforms the state-of-the-art review-based recommender systems on 7 real-world datasets. Qualitative analyses also show that LUAR can accurately focus on important auxiliary reviews.
ISSN:0020-0255
1872-6291
DOI:10.1016/j.ins.2020.09.025