Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hi-Gen: Generative Retrieval For Large-Scale Personalized E-commerce Search

Yanjing Wu, Yinfu Feng, Jian Wang, Wenji Zhou, Yunan Ye, Rong Xiao, Jun Xiao

TL;DR

Hi-Gen addresses key gaps in generative retrieval for large-scale personalized e-commerce search by jointly encoding semantic relevance and item efficiency into docIDs and applying a position-aware decoding loss. The method combines discriminative representation learning with a category-guided hierarchical clustering strategy to generate meaningful docIDs, and introduces a LM trained with a loss that emphasizes earlier, semantically relevant, and efficiently clustered tokens. Two online recall variants further enable real-time large-scale retrieval, and extensive experiments on AOL4PS and AEDST, plus online AB tests, show state-of-the-art Recall@k improvements and positive business impact. The work demonstrates practical viability of gen­erative retrieval in industry-scale search, including strong zero-shot performance and scalable online recall.

Abstract

Leveraging generative retrieval (GR) techniques to enhance search systems is an emerging methodology that has shown promising results in recent years. In GR, a text-to-text model maps string queries directly to relevant document identifiers (docIDs), dramatically simplifying the retrieval process. However, when applying most GR models in large-scale E-commerce for personalized item search, we must face two key problems in encoding and decoding. (1) Existing docID generation methods ignore the encoding of efficiency information, which is critical in E-commerce. (2) The positional information is important in decoding docIDs, while prior studies have not adequately discriminated the significance of positional information or well exploited the inherent interrelation among these positions. To overcome these problems, we introduce an efficient Hierarchical encoding-decoding Generative retrieval method (Hi-Gen) for large-scale personalized E-commerce search systems. Specifically, we first design a representation learning model using metric learning to learn discriminative feature representations of items to capture semantic relevance and efficiency information. Then, we propose a category-guided hierarchical clustering scheme that makes full use of the semantic and efficiency information of items to facilitate docID generation. Finally, we design a position-aware loss to discriminate the importance of positions and mine the inherent interrelation between different tokens at the same position. This loss boosts the performance of the language model used in the decoding stage. Besides, we propose two variants of Hi-Gen (Hi-Gen-I2I and Hi-Gen-Cluster) to support online real-time large-scale recall in the online serving process. Hi-Gen gets 3.30% and 4.62% improvements over SOTA for Recall@1 on the public and industry datasets, respectively.

Hi-Gen: Generative Retrieval For Large-Scale Personalized E-commerce Search

TL;DR

Hi-Gen addresses key gaps in generative retrieval for large-scale personalized e-commerce search by jointly encoding semantic relevance and item efficiency into docIDs and applying a position-aware decoding loss. The method combines discriminative representation learning with a category-guided hierarchical clustering strategy to generate meaningful docIDs, and introduces a LM trained with a loss that emphasizes earlier, semantically relevant, and efficiently clustered tokens. Two online recall variants further enable real-time large-scale retrieval, and extensive experiments on AOL4PS and AEDST, plus online AB tests, show state-of-the-art Recall@k improvements and positive business impact. The work demonstrates practical viability of gen­erative retrieval in industry-scale search, including strong zero-shot performance and scalable online recall.

Abstract

Leveraging generative retrieval (GR) techniques to enhance search systems is an emerging methodology that has shown promising results in recent years. In GR, a text-to-text model maps string queries directly to relevant document identifiers (docIDs), dramatically simplifying the retrieval process. However, when applying most GR models in large-scale E-commerce for personalized item search, we must face two key problems in encoding and decoding. (1) Existing docID generation methods ignore the encoding of efficiency information, which is critical in E-commerce. (2) The positional information is important in decoding docIDs, while prior studies have not adequately discriminated the significance of positional information or well exploited the inherent interrelation among these positions. To overcome these problems, we introduce an efficient Hierarchical encoding-decoding Generative retrieval method (Hi-Gen) for large-scale personalized E-commerce search systems. Specifically, we first design a representation learning model using metric learning to learn discriminative feature representations of items to capture semantic relevance and efficiency information. Then, we propose a category-guided hierarchical clustering scheme that makes full use of the semantic and efficiency information of items to facilitate docID generation. Finally, we design a position-aware loss to discriminate the importance of positions and mine the inherent interrelation between different tokens at the same position. This loss boosts the performance of the language model used in the decoding stage. Besides, we propose two variants of Hi-Gen (Hi-Gen-I2I and Hi-Gen-Cluster) to support online real-time large-scale recall in the online serving process. Hi-Gen gets 3.30% and 4.62% improvements over SOTA for Recall@1 on the public and industry datasets, respectively.
Paper Structure (22 sections, 7 equations, 4 figures, 6 tables, 1 algorithm)

This paper contains 22 sections, 7 equations, 4 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Methodology
  4. Overview
  5. Discriminative Representation Learning
  6. Atomic Embedding Learning Module
  7. Metric Learning Module
  8. Document Identifiers (docIDs) Generation
  9. Language Model With Position-aware Loss
  10. Online Serving Optimization
  11. Experiments Setup
  12. Datasets
  13. Baselines
  14. Implementation Details
  15. Evaluation Metrics
  16. ...and 7 more sections

Figures (4)

  • Figure 1: Overview of the proposed Hi-Gen. In the figure, the arrow indicates the data flow from input to output.
  • Figure 2: Overview of discriminative representation learning.
  • Figure 3: An illustration of the category-guided hierarchical clustering method that produced the docID.
  • Figure 4: Recall@5000 and RecallNum for different variants of Hi-Gen in AEDST.