Connecting user and item perspectives in popularity debiasing for collaborative recommendation

作者：

Highlights：

• We propose two new metrics for monitoring popularity bias in recommendation.

• Pair- and point-wise optimizations emphasize popularity-biased recommendations.

• We propose a mitigation procedure based on a new data sampling and regularization.

• Treated models are less biased on popularity and better meet beyond-accuracy goals.

摘要

•We propose two new metrics for monitoring popularity bias in recommendation.•Pair- and point-wise optimizations emphasize popularity-biased recommendations.•We propose a mitigation procedure based on a new data sampling and regularization.•Treated models are less biased on popularity and better meet beyond-accuracy goals.

论文关键词：Recommender systems,Popularity bias,Beyond-accuracy

论文评审过程：Received 16 November 2019, Revised 9 September 2020, Accepted 9 September 2020, Available online 29 September 2020, Version of Record 29 September 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2020.102387