PD-SRS: Personalized Diversity for a Fair Session-Based Recommendation System
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Session-based Recommender Systems (SRSs), which aim to recommend users’ next action based on their current and historical sessions, play a significant role in many real-world online services. The existing session-based recommendation methods have mainly focused on the accuracy of recommendation, which biases to reinforce popular items/services and loses the recommendation diversity. Diversity is a positive aspect particularly in SRSs as the target user may like to be surprised and interact with a broader range of content in different sessions. In this work, we propose a Personalized Diversification strategy for a Session-based Recommender System (PD-SRS) using graph neural networks. Comprehensive experiments are carried out on two real-world datasets to demonstrate the effectiveness of PD-SRS in making a trade-off between accuracy and personalized diversity over the baselines.
Open Access Status
This publication is not available as open access