PURS: Personalized Unexpected Recommender System for Improving User Satisfaction

TL;DR

This paper introduces PURS, a personalized recommender system designed to enhance user satisfaction by balancing accuracy with unexpectedness, effectively addressing the filter bubble problem through multi-cluster user interest modeling and self-attention mechanisms.

Contribution

The paper presents a novel personalized unexpected recommender system that integrates multi-cluster interest modeling and self-attention to improve recommendation diversity and user engagement.

Findings

PURS outperforms baseline methods in accuracy and unexpectedness.

Online A/B testing shows a 3% increase in average video views per user.

The model is being deployed by Alibaba-Youku.

Abstract

Classical recommender system methods typically face the filter bubble problem when users only receive recommendations of their familiar items, making them bored and dissatisfied. To address the filter bubble problem, unexpected recommendations have been proposed to recommend items significantly deviating from user's prior expectations and thus surprising them by presenting "fresh" and previously unexplored items to the users. In this paper, we describe a novel Personalized Unexpected Recommender System (PURS) model that incorporates unexpectedness into the recommendation process by providing multi-cluster modeling of user interests in the latent space and personalized unexpectedness via the self-attention mechanism and via selection of an appropriate unexpected activation function. Extensive offline experiments on three real-world datasets illustrate that the proposed PURS model significantly outperforms the state-of-the-art baseline approaches in terms of both accuracy and unexpectedness measures. In addition, we conduct an online A/B test at a major video platform Alibaba-Youku, where our model achieves over 3\% increase in the average video view per user metric. The proposed model is in the process of being deployed by the company.