A Bird's-eye View of Reranking: from List Level to Page Level

TL;DR

This paper introduces a novel page-level reranking model that captures inter-list interactions and page format effects, significantly improving recommendation performance in multi-list settings.

Contribution

The paper proposes the Page-level Attentional Reranking (PAR) model with hierarchical attention, spatial-aware influence modeling, and a mixture-of-experts to better capture user behaviors across multiple lists.

Findings

PAR outperforms baseline models on public and proprietary datasets.

The hierarchical attention effectively models intra- and inter-list interactions.

Spatial-scaled attention improves modeling of page format effects.

Abstract

Reranking, as the final stage of multi-stage recommender systems, refines the initial lists to maximize the total utility. With the development of multimedia and user interface design, the recommendation page has evolved to a multi-list style. Separately employing traditional list-level reranking methods for different lists overlooks the inter-list interactions and the effect of different page formats, thus yielding suboptimal reranking performance. Moreover, simply applying a shared network for all the lists fails to capture the commonalities and distinctions in user behaviors on different lists. To this end, we propose to draw a bird's-eye view of \textbf{page-level reranking} and design a novel Page-level Attentional Reranking (PAR) model. We introduce a hierarchical dual-side attention module to extract personalized intra- and inter-list interactions. A spatial-scaled attention network is devised to integrate the spatial relationship into pairwise item influences, which explicitly models the page format. The multi-gated mixture-of-experts module is further applied to capture the commonalities and differences of user behaviors between different lists. Extensive experiments on a public dataset and a proprietary dataset show that PAR significantly outperforms existing baseline models.