Simple but Efficient: A Multi-Scenario Nearline Retrieval Framework for Recommendation on Taobao

In recommendation systems, the matching stage is becoming increasingly critical, serving as the upper limit for the entire recommendation process. Recently, some studies have started to explore the use of multi-scenario information for recommendations, such as model-based and data-based approaches. However, the matching stage faces significant challenges due to the need for ultra-large-scale retrieval and meeting low latency requirements. As a result, the methods applied at this stage (collaborative filtering and two-tower models) are often designed to be lightweight, hindering the full utilization of extensive information. On the other hand, the ranking stage features the most sophisticated models with the strongest scoring capabilities, but due to the limited screen size of mobile devices, most of the ranked results may not gain exposure or be displayed. In this paper, we introduce an innovative multi-scenario nearline retrieval framework. It operates by harnessing ranking logs from various scenarios through Flink, allowing us to incorporate finely ranked results from other scenarios into our matching stage in near real-time. Besides, we propose a streaming scoring module, which selects a crucial subset from the candidate pool. Implemented on the "Guess You Like" (homepage of the Taobao APP), China's premier e-commerce platform, our method has shown substantial improvements-most notably, a 5% uptick in product transactions. Furthermore, the proposed approach is not only model-free but also highly efficient, suggesting it can be quickly implemented in diverse scenarios and demonstrate promising performance.

Graph Contrastive Learning with Multi-Objective for Personalized Product Retrieval in Taobao Search

In e-commerce search, personalized retrieval is a crucial technique for improving user shopping experience. Recent works in this domain have achieved significant improvements by the representation learning paradigm, e.g., embedding-based retrieval (EBR) and collaborative filtering (CF). EBR methods do not sufficiently exploit the useful collaborative signal and are difficult to learn the representations of long-tail item well. Graph-based CF methods improve personalization by modeling collaborative signal within the user click graph. However, existing Graph-based methods ignore user's multiple behaviours, such as click/purchase and the relevance constraint between user behaviours and items.In this paper, we propose a Graph Contrastive Learning with Multi-Objective (GCL-MO) collaborative filtering model, which solves the problems of weak relevance and incomplete personalization in e-commerce search. Specifically, GCL-MO builds a homogeneous graph of items and then optimizes a multi-objective function of personalization and relevance. Moreover, we propose a modified contrastive loss for multi-objectives graph learning, which avoids the mutual suppression among positive samples and thus improves the generalization and robustness of long-tail item representations. These learned item embeddings are then used for personalized retrieval by constructing an efficient offline-to-online inverted table. GCL-MO outperforms the online collaborative filtering baseline in both offline/online experimental metrics and shows a significant improvement in the online A/B testing of Taobao search.

Rethinking the Role of Pre-ranking in Large-scale E-Commerce Searching System

E-commerce search systems such as Taobao Search, the largest e-commerce searching system in China, aim at providing users with the most preferred items (e.g., products). Due to the massive data and limited time for response, a typical industrial ranking system consists of three or more modules, including matching, pre-ranking, and ranking. The pre-ranking is widely considered a mini-ranking module, as it needs to rank hundreds of times more items than the ranking under limited latency. Existing researches focus on building a lighter model that imitates the ranking model. As such, the metric of a pre-ranking model follows the ranking model using Area Under ROC (AUC) for offline evaluation. However, such a metric is inconsistent with online A/B tests in practice, so engineers have to perform costly online tests to reach a convincing conclusion. In our work, we rethink the role of the pre-ranking. We argue that the primary goal of the pre-ranking stage is to return an optimal unordered set rather than an ordered list of items because it is the ranking that determines the final exposures. Since AUC measures the quality of an ordered item list, it is not suitable for evaluating the quality of the output unordered set. This paper proposes a new evaluation metric called All-Scenario Hitrate (ASH) for pre-ranking. ASH is proven effective in the offline evaluation and consistent with online A/B tests based on numerous experiments in Taobao Search. We also introduce an all-scenario-based multi-objective learning framework (ASMOL), which improves the ASH significantly. Surprisingly, the new pre-ranking model can outperforms the ranking model when outputting thousands of items. The phenomenon validates that the pre-ranking stage should not imitate the ranking blindly. With the improvements in ASH consistently translating to online improvement, it makes a 1.2% GMV improvement on Taobao Search.