Event-based Product Carousel Recommendation with Query-Click Graph

TL;DR

This work tackles event-based product carousel recommendations, where an event like Valentine’s Day comprises multiple shopping aspects and lacks ground-truth mappings to aspect-specific products. It introduces a recommender that builds a query-click bipartite graph over event-related queries and product-types, then learns event aspects via iterative clustering on the graph, forming carousels for each aspect and ranking items by $CTR(a,p) = \frac{\sum_{q \in a} Clk_{(q,p)}}{\sum_{q \\in a} Imp_{(q,p)}}$. Key contributions include formalizing the problem, proposing the iterative clustering approach that links query signals to product-types, and demonstrating improved precision with coherent, diverse carousels across multiple events, including case studies. This method enables automatic, scalable generation of multi-aspect event carousels without manual curation, supporting dynamic catalogs and evolving events in practical e-commerce settings.

Abstract

Many current recommender systems mainly focus on the product-to-product recommendations and user-to-product recommendations even during the time of events rather than modeling the typical recommendations for the target event (e.g., festivals, seasonal activities, or social activities) without addressing the multiple aspects of the shopping demands for the target event. Product recommendations for the multiple aspects of the target event are usually generated by human curators who manually identify the aspects and select a list of aspect-related products (i.e., product carousel) for each aspect as recommendations. However, building a recommender system with machine learning is non-trivial due to the lack of both the ground truth of event-related aspects and the aspect-related products. To fill this gap, we define the novel problem as the event-based product carousel recommendations in e-commerce and propose an effective recommender system based on the query-click bipartite graph. We apply the iterative clustering algorithm over the query-click bipartite graph and infer the event-related aspects by the clusters of queries. The aspect-related recommendations are powered by the click-through rate of products regarding each aspect. We show through experiments that this approach effectively mines product carousels for the target event.

Figures (2)

• Figure 1: Product Carousels for Labor Day on Walmart.com. Each carousel, presented by its icon in (a), represents a event-related aspect and users can browse the product carousel of a certain aspect by a click on the carousel's icon, for example, the patio carousel in (b) and the electronics carousel in (c).
• Figure 2: (a) the training stage of our event-based product carousel recommdner system. Given an event, we create the bipartite graph and apply the iterative clustering algorithm to infer the aspects and corresponding product pools. (b) the recommendation stage. Given an event, our recommender system uses the inferred aspects and for each aspect, we rank the product in the corresponding product pool by the click-through rate in descending order as the product carousel for this aspect.

Theorems & Definitions (1)

• Definition 3.1