Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimizing E-commerce Search: Toward a Generalizable and Rank-Consistent Pre-Ranking Model

Enqiang Xu, Yiming Qiu, Junyang Bai, Ping Zhang, Dadong Miao, Songlin Wang, Guoyu Tang, Lin Liu, Mingming Li

TL;DR

The paper tackles pre-ranking in large-scale e-commerce search by introducing GRACE, a model that jointly optimizes rank consistency and generalization. It adds a rank-consistency objective derived from ranking positions and a contrastive generalization objective using hash-ID and attribute embeddings aligned with pre-trained signals via Info-NCE. Empirical results show GRACE achieves notable offline gains in AUC (+0.749% over PLE) and online improvements in CVR (+1.28%) and GMV (+1.62%), with stronger benefits on long-tail items. The approach is lightweight for deployment, scalable to billions of items, and demonstrates practical impact for industrial search systems.

Abstract

In large e-commerce platforms, search systems are typically composed of a series of modules, including recall, pre-ranking, and ranking phases. The pre-ranking phase, serving as a lightweight module, is crucial for filtering out the bulk of products in advance for the downstream ranking module. Industrial efforts on optimizing the pre-ranking model have predominantly focused on enhancing ranking consistency, model structure, and generalization towards long-tail items. Beyond these optimizations, meeting the system performance requirements presents a significant challenge. Contrasting with existing industry works, we propose a novel method: a Generalizable and RAnk-ConsistEnt Pre-Ranking Model (GRACE), which achieves: 1) Ranking consistency by introducing multiple binary classification tasks that predict whether a product is within the top-k results as estimated by the ranking model, which facilitates the addition of learning objectives on common point-wise ranking models; 2) Generalizability through contrastive learning of representation for all products by pre-training on a subset of ranking product embeddings; 3) Ease of implementation in feature construction and online deployment. Our extensive experiments demonstrate significant improvements in both offline metrics and online A/B test: a 0.75% increase in AUC and a 1.28% increase in CVR.

Optimizing E-commerce Search: Toward a Generalizable and Rank-Consistent Pre-Ranking Model

TL;DR

The paper tackles pre-ranking in large-scale e-commerce search by introducing GRACE, a model that jointly optimizes rank consistency and generalization. It adds a rank-consistency objective derived from ranking positions and a contrastive generalization objective using hash-ID and attribute embeddings aligned with pre-trained signals via Info-NCE. Empirical results show GRACE achieves notable offline gains in AUC (+0.749% over PLE) and online improvements in CVR (+1.28%) and GMV (+1.62%), with stronger benefits on long-tail items. The approach is lightweight for deployment, scalable to billions of items, and demonstrates practical impact for industrial search systems.

Abstract

In large e-commerce platforms, search systems are typically composed of a series of modules, including recall, pre-ranking, and ranking phases. The pre-ranking phase, serving as a lightweight module, is crucial for filtering out the bulk of products in advance for the downstream ranking module. Industrial efforts on optimizing the pre-ranking model have predominantly focused on enhancing ranking consistency, model structure, and generalization towards long-tail items. Beyond these optimizations, meeting the system performance requirements presents a significant challenge. Contrasting with existing industry works, we propose a novel method: a Generalizable and RAnk-ConsistEnt Pre-Ranking Model (GRACE), which achieves: 1) Ranking consistency by introducing multiple binary classification tasks that predict whether a product is within the top-k results as estimated by the ranking model, which facilitates the addition of learning objectives on common point-wise ranking models; 2) Generalizability through contrastive learning of representation for all products by pre-training on a subset of ranking product embeddings; 3) Ease of implementation in feature construction and online deployment. Our extensive experiments demonstrate significant improvements in both offline metrics and online A/B test: a 0.75% increase in AUC and a 1.28% increase in CVR.
Paper Structure (20 sections, 11 equations, 1 figure, 4 tables)

This paper contains 20 sections, 11 equations, 1 figure, 4 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Task Definition
  4. Base Model
  5. Rank Consistency Loss
  6. Generalization Loss
  7. Total Loss
  8. Experiments
  9. Setup
  10. Dataset
  11. Metrics
  12. Baseline models
  13. Hyperparameter
  14. Comparison with baseline models
  15. Ablation Study
  16. ...and 5 more sections

Figures (1)

  • Figure 1: The proposed framework, where "Attrs" refers to attribute information such as brand, shop, and category.