Abstract:Session-based recommendation leverages the short-term interaction sequence to predict the next item a user is most likely to click on. Generally, in real applications, users often click on different types of items in the same session, which makes the items in the sequence present diversity, and degree of diversity vary with different sequences, that is, there is a phenomenon of personalized diversity. However, the majority of existing session recommendation models primarily focus on improving recommendation accuracy and overlooking the importance of personalized diversity, which leads to the issue of filtering bubbles. In order to alleviate the long-tail effect caused by the filtering bubble, the recommendation system should not only consider the accuracy of recommended items, but also the diversity of users' demands for recommended items. To address this issue, this paper proposes a model, named as LAP-SR, a L ong-tail A lleviation P ost-processing framework based on personalized diversity of S ession R ecommendation. After obtaining the initial recommendation list through the selected initial session model, LAP-SR model generates the long-tail item set and item graph according to session contents. Besides, from the item graph, the model calculates the embedding representation of items, from which the diversity of all sessions is calculated. By combining the obtained diversity and the long-tail item set, the initial recommendation list is soft processed. On this basis, the model yields the final recommendation list, which alleviates the overall long-tail effect. We verified that as a post-processing framework, the proposed model can be applied to various session recommendation models. Extensive experiments on the three datasets of Diginetica, Tmall and Yoochoose1_64 demonstrate that the LAP-SR model has a competitive advantage over the baseline models in mitigating the long-tail effect.

A Simple Yet Effective Approach for Diversified Session-Based Recommendation

Covering Diversification and Fairness for Better Recommendation (Short Paper).

Understanding Diversity in Session-Based Recommendation

Purpose Tendency-Aware Diversified Strategy for Effective Session-Based Recommendation.

Diversifying Personalized Recommendation with User-session Context

On Practical Diversified Recommendation with Controllable Category Diversity Framework

Personalizing Diversity Versus Accuracy in Session-Based Recommender Systems

Feature-aware Diversified Re-ranking with Disentangled Representations for Relevant Recommendation

Disentangling Interest and Conformity for Eliminating Popularity Bias in Session-Based Recommendation

Improving End-to-End Sequential Recommendations with Intent-aware Diversification

A long-tail alleviation post-processing framework based on personalized diversity of session recommendation

A Joint Optimization Approach for Personalized Recommendation Diversification.

Sequential/Session-based Recommendations: Challenges, Approaches, Applications and Opportunities

CARE: Context-aware Attention Interest Redistribution for Session-Based Recommendation

Recent Advances in Diversified Recommendation

Solving the Apparent Diversity-Accuracy Dilemma of Recommender Systems

Diversify or Not: Dynamic Diversification for Personalized Recommendation

A Survey on Session-based Recommender Systems

A joint optimization approach for personalized recommendation diversification

A Novel Framework for Improving Recommender Diversity.

Relevance Meets Coverage: A Unified Framework to Generate Diversified Recommendations.