Balancing the trade-off between accuracy and diversity in recommender systems with personalized explanations based on Linked Open Data

作者：

Highlights：

• Formalization of the popularity bias and transparency problems in collaborative filtering recommendation algorithms.

• Proposal of personalized extraction of the users’ most relevant Knowledge Graph properties.

• Proposal of an explainable multi-domain item reordering system based on the best explanation for recommendations.

• Evaluation of proposal comparing accuracy and beyond-accuracy metrics such as diversity, transparency, fairness, and coverage.

摘要

•Formalization of the popularity bias and transparency problems in collaborative filtering recommendation algorithms.•Proposal of personalized extraction of the users’ most relevant Knowledge Graph properties.•Proposal of an explainable multi-domain item reordering system based on the best explanation for recommendations.•Evaluation of proposal comparing accuracy and beyond-accuracy metrics such as diversity, transparency, fairness, and coverage.

论文关键词：Recommender systems,Collaborative filtering,Linked open data,Explainable AI

论文评审过程：Received 31 October 2021, Revised 23 June 2022, Accepted 24 June 2022, Available online 30 June 2022, Version of Record 12 July 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109333